MYC Messenger RNA Expression Predicts Survival Outcome in Childhood Primitive Neuroectodermal Tumor/Medulloblastoma¹

CNS³ PNET, including PNET/MB, are the most common malignant brain tumors in children and constitute 20–25% of all pediatric brain tumors (1). Because of the high risk of leptomeningeal dissemination, standard postoperative treatment for PNETs includes not only local radiotherapy but also craniospinal radiotherapy and chemotherapy (2, 3). Such treatment causes long-term morbidity including endocrine and growth disturbances, as well as neurocognitive dysfunction, which is particularly severe in young children (4, 5, 6, 7, 8, 9).

One strategy to reduce treatment-related long-term effects in patients at the greatest risk for severe radiation-induced brain injury is to delay radiotherapy through the use of prolonged chemotherapy (10, 11). Nearly two decades ago, van Eys et al. (12) used multiagent chemotherapy including mechlorethamine, vincristine, procarbazine, and prednisone as primary therapy after surgery for infants with malignant brain tumors. Only those patients with treatment failure and tumor progression received radiotherapy. Six of 12 infants with PNET/MB treated between 1976 and 1988 remain tumor free without having received radiotherapy and have been followed for a median time of 10.6 years (range, 6.2–15.2 years; Ref. 13). Cognitive functions in this group were within the normal ranges, stable across annual assessments, and clearly better than those in children who received radiation therapy. In older PNET/MB patients, one strategy to reduce treatment-related long-term effects is to reduce craniospinal radiation doses. Goldwein et al. (14) used 18 Gy of radiation to the craniospinal axis, a posterior fossa boost to 50.4–55.8 Gy, and chemotherapy to treat PNET/MB patients between 18 and 60 months of age. Seven of 10 patients survived with a median follow-up of 6.3 years, and the mean intelligence quotient of these patients was within the normal ranges.

Clinical prognostic factors for PNETs have been identified over the last two decades (15, 16, 17, 18, 19). These include metastatic stage, extent of tumor resection, tumor location, and age and are currently used to distinguish a high-risk group of patients (metastatic stage ≥ 1, residual tumor bulk ≥ 1.5 cm², age < 3 years, supratentorial tumor location) from a standard-risk group. However, investigators have appropriately avoided the use of the term “good risk” or “good outcome” because published survival rates in the 50–60% range do not justify this term. At the present time, clinical prognostic factors do not identify a “good outcome” group of PNET/MB patients that can be treated with a substantially less toxic treatment strategy compared with standard treatment. Because it is unlikely that new clinical prognostic factors will be found, biological prognostic factors must be identified to further improve PNET/MB prognostic systems.

Several biological factors have been studied as candidate prognostic factors in PNETs. Glial fibrillary acidic protein (GFAP) expression (20), HER2/HER4 coexpression (21), Ki-67 (MIB-1) proliferation index (22), and mitotic index (23) have been analyzed in multivariable analysis and found to have independent predictive value. When hazard ratios were reported, the prognostic impact of these factors was similar or less than that of clinical factors. By contrast, the predictive value of TrkC mRNA expression has been shown to exceed that of all clinical factors (24, 25).

MYC (c-myc) has been the subject of several published studies directed at identifying biological prognostic markers in human PNET/MB. Most show that MYC gene amplification is uncommon in PNET/MB, with an incidence of ∼8% in primary tumors (26, 27, 28, 29, 30, 31). The incidence of MYC gene amplification in PNET/MB cell lines and xenografts may be higher and ranges up to 67% (32, 33), suggesting that MYC gene amplification correlates with cell line establishment and tumorigenicity. MYC gene amplification has been suggested as an indicator of poor prognosis in case reports (29, 34, 35) and in a recently published study of 29 PNET/MB patients (26). In aggregate, however, published studies suggest that the frequency of MYC amplification is not sufficiently high to provide prognostic information of greater value than clinical variables.

Mechanisms to activate MYC other than gene amplification are well recognized in various solid tumors. They include retroviral insertional mutagenesis, chromosomal translocation, somatic mutations, or activation by transcription factors (reviewed in Ref. 36). Transcriptional regulation of MYC may involve the adenomatous polyposis coli (APC) and β-catenin (CTNNB1) pathways, which may be involved in PNET pathogenesis. A subset of PNETs is associated with Turcot’s syndrome (37), which is the association of colonic cancer with primary brain tumors and is characterized by germ-line mutations in the APC gene. Whereas mutations of APC have not been detected in sporadic PNETs (38, 39), approximately 5% of sporadic PNETs have been reported to contain mutations in a second Wingless/Wnt pathway member, β-catenin (40, 41). Mutations in either APC or β-catenin act to stabilize β-catenin protein. Stabilized β-catenin protein accumulates and translocates into the nucleus, where it complexes with Tcf4 and up-regulates the transcription of MYC as well as other targets (42, 43). In a recent study, aberrant nuclear β-catenin staining has been demonstrated in 18% of sporadic PNET/MB and in one PNET/MB from patient with Turcot’s syndrome (41). Therefore, mechanisms other than gene amplification may activate MYC transcription in PNET/MB. However, expression levels of MYC transcripts have not been reported for a representative group of PNET/MB.

To study the biological significance of MYC in PNET, we measured MYC mRNA expression levels in 6 PNET cell lines and 26 primary PNET/MB by semiquantitative RT-PCR and compared MYC mRNA expression levels with MYC gene copy number, clinical and biological variables, and survival outcomes.

Seventy-eight patients were diagnosed with a PNET/MB at the Children’s Hospital of Philadelphia between January 1988 and December 1998. All diagnoses were confirmed by histological assessment of a tumor specimen obtained at surgery by one neuropathologist (L. B. R.). Adequate frozen tumor tissue to perform RT-PCR was available from 26 PNET/MB patients. Tumor samples were snap-frozen in liquid nitrogen in the operating room and then stored at −80°C until further analysis. The demographic and treatment characteristics of the 26 study patients are summarized in Table 1. They are comparable with those of the 52 PNET/MB patients not included in the present study due to lack of frozen tumor tissue. Therefore, the subset of patients included in the present study can be considered representative. The median age at diagnosis for these PNET/MB patients was 6.4 years (range, 0.3–14.8 years). Seventeen (65%) PNET/MB patients were male, and nine (35%) were female. Evidence of leptomeningeal metastasis (M_1–3) was present in 7 (27%) PNET/MB patients, and 19 (73%) patients were M₀. Tumor location was cerebellar in all patients. Twenty (77%) patients had ≥90% surgical tumor resection, six (23%) patients had 50–90% resection of the tumor, and no patient had <50% of the tumor resected. Postoperative therapy included radiation and/or chemotherapy. Twenty patients had combined radiotherapy and chemotherapy, two patients had radiotherapy alone, and four infants had chemotherapy alone. Chemotherapy was administered to 15 patients according to a previously described protocol including vincristine, lomustine, and cisplatin (44), to 7 younger children according to infant brain tumor protocols (10), and to 2 patients according to other regimens. Median follow-up for the 15 patients who remain alive and progression free at the time of this report was 62 months (range, 15–105 months). Approval to link laboratory data to clinical data has been obtained by the Institutional Review Board.

DAOY and PFSK human PNET cells were purchased from American Type Culture Collection (Manassas, VA). D341, D425, and D458 human PNET cells were a kind gift from Dr. Henry Friedman (Duke University, Durham, NC). UW228-2 human PNET cells were a kind gift from Dr. John R. Silber (University of Washington, Seattle, WA). DAOY, D341, D425, and D458 cells were cultured in Richter’s Zinc Option medium/10% FBS; PFSK cells were cultured in RPMI 1640/10% FBS; UW228-2 cells were cultured in DMEM/10% FBS. All cell cultures were maintained at 37°C in a humidified atmosphere with 5% CO₂.

Tumor tissue was disrupted with a sterile disposable tissue grinder (Sage Products Inc., Crystal Lake, IL). After tissue homogenization in guanidinium isothiocyanate-containing buffer, total RNA was isolated using a silica gel-based method according to procedures recommended by the manufacturer (RNeasy; Qiagen Inc., Valencia, CA). This included DNase digestion of the purified RNA. The yield of RNA from the various samples was calculated by spectrophotometry. RT reactions were carried out using 1 μg of total RNA in a total volume of 20 μl containing 150 ng of random hexamers, 0.5 mm deoxynucleotide triphosphates, 10 mm DTT, and 200 units of SuperScript II reverse transcriptase (all from Life Technologies, Inc., Gaithersburg, MD) in 20 mm Tris-HCl (pH 8.4) with 50 mm KCl and 2.5 mm MgCl₂. Total RNA was denatured at 70°C for 10 min and immediately chilled on ice. First-strand cDNAs were obtained after 10 min at 23°C and 50 min at 42°C. The reaction was terminated at 70°C (15 min). Two units of RNase H (Life Technologies, Inc.) were added to each RT reaction followed by incubation at 37°C for 20 min.

The method used for semiquantitative RT-PCR has been described previously (45, 46). In brief, PCR was carried out in a final volume of 20 μl containing 1 unit of Taq Gold polymerase, 100 μm deoxynucleotide triphosphates, 0.4 μm of each primer, buffer consisting of 50 mm KCl and 10 mm Tris-HCl (pH 8.3), 1.5 mm MgCl₂, and 2 μl of the RT product as template. PCR primers were designed to bracket cDNA sequences that cross an intron-exon boundary in genomic DNA. Primer sequences specific for MYC (c-myc) were 5′-TGGTCTTCCCCTACCCTCTCAAC-3′ (sense) and 5′-GATCCAGACTCTGACCTTTTGCC-3′ (antisense) with an expected PCR product size of 265 bp. MYC primers were biotinylated at their 5′ ends. Semiquantitation was accomplished by comparison to signal generated by multiplexed primers to amplify GAPD in each reaction (sense, 5′-CATCAAGAAGGTGGTGAAGC-3′; antisense, 5′-GAGCTTGACAAAGTGGTCGT-3′; product size, 160 bp). Due to the high expression of GAPD, primers were mixed at a ratio of 1:199 biotinylated:unbiotinylated to ensure that all signals remained within a linear dynamic range. Amplification was performed on a PTC-100 Programmable Thermal Controller (MJ Research Inc., Waltham, MA). The samples were initially denatured at 95°C for 12 min, followed by 18 cycles with denaturation at 95°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 90 s. The final cycle was followed by a 5-min extension step at 72°C. The absence of contaminants was routinely checked by RT-PCR assays of negative control samples (H₂O control or no reverse transcriptase added). Ten μl of each PCR sample were analyzed in parallel with a biotinylated molecular weight marker (Amersham, Arlington Heights, IL) on a nondenaturing 6% polyacrylamide gel. DNA was electrotransferred to a nylon membrane and immobilized by UV cross-linking. Detection of biotin-labeled DNA was performed by a chemiluminescent detection system (Tropix, Bedford, MA). To provide a semiquantitative assessment of RT-PCR-studies, we performed densitometric analysis of transcript signals using the NIH Image program (United States NIH).⁴ Expression levels of the target transcripts were normalized by the use of the internal standard GAPD. The analysis of RT-PCR was blind to all clinical data.

Frozen tumor tissue adequate to isolate genomic DNA and perform PCR was available from 11 PNET/MB patients. Genomic DNA was isolated using the QIAmp DNA Mini kit (Qiagen Inc.) according to procedures recommended by the manufacturer. In brief, tumor tissue was disrupted with a sterile disposable tissue grinder (Sage Products Inc.), lysed with proteinase K, and incubated with RNase A. Genomic DNA was then purified using a silica gel-based method. The yield of DNA from the various samples was calculated by spectrophotometry. The method used for semiquantitative PCR was similar to that used for RT-PCR described above, with the following difference: biotinylated GAPD primers were used instead of a 1:199 biotinylated:unbiotinylated GAPD primer mixture. Moreover, the cycle number was 20 instead of 18. Cell lines and tumors with MYC/GAPD signal ratios of >1.5 were considered to be indicative of MYC gene amplification, as confirmed through analyses of cell lines of known MYC copy number.

TrkC mRNA expression, assessed by in situ hybridization, was available for 19 of the 26 patients in the present study and has been published previously (25). GFAP expression, assessed by immunohistochemistry, was published previously (20). Information on seven additional cases has been added in the present study.

Clinical characteristics of included and excluded patients were compared using Fisher’s exact tests. Relative risk of progression or death was calculated by univariate and multivariable analysis using Cox regression models (47). Progression-free survival and overall survival were determined by Kaplan-Meier analysis (48), and differences between survival curves were calculated using the Mantel log-rank test (49) to confirm the results based on Cox regression models.

PNET cell lines showed a wide range of MYC mRNA expression, with the lowest expression in DAOY PNET cells (normalized expression level = 0.85) and the highest expression in D425 PNET cells (normalized expression level = 19.33; Fig. 1 A). These results, which are consistent with Northern blot results published previously by others (50), validate the methods used in this study. We also used semiquantitative PCR to identify MYC gene amplification. D425, D341, and D458 PNET cells had high MYC:GAPD signal ratios (4.6, 4.1, and 2.4, respectively), indicating MYC gene amplification. UW228-2, PFSK, and DAOY had low MYC:GAPD signals (0.27, 0.38, and 0.54, respectively), indicating single copy numbers. These results are consistent with Southern blot results for these cell lines published previously (33, 50). Moreover, we were able to detect 3-fold MYC gene amplification in the PNET cell line D283 (data not shown), indicating a high sensitivity of the assay. In DAOY, PFSK, and UW288-2 PNET cell lines that do not have MYC gene amplification, MYC mRNA expression levels showed a 10-fold difference, indicating that mechanisms other than MYC gene amplification can contribute to increased MYC mRNA expression.

PNET/MB primary tumors showed also a large range of MYC mRNA expression. The median normalized MYC mRNA expression level in primary PNET/MB was 2.5 (range, 0.5–10.6) with a 22-fold difference between the highest value and the lowest value (Fig. 1 B). Log-transformed MYC mRNA expression levels showed a normal distribution pattern. Tumors with MYC mRNA expression ≥ the median (i.e., 2.5) were defined as PNET/MB with high MYC mRNA expression, whereas those with MYC mRNA expression levels <2.5 were defined as PNET/MB with low MYC mRNA expression.

To compare MYC mRNA expression levels with MYC gene amplification in primary PNET/MB, we selected 11 PNET/MB with sufficient frozen tissue to isolate genomic DNA and performed PCR. We found that MYC gene amplification was present in 3 of 11 PNET/MB. Similar to the PNET cell lines, MYC mRNA expression was highly variable in the PNET/MB with no MYC gene amplification and showed an 11.6-fold difference between the highest and lowest values (Fig. 1 B).

Table 2 summarizes MYC mRNA expression, TrkC mRNA expression, GFAP expression, patient characteristics, treatment, and survival outcomes. Ten of 26 patients died as a result of progressive disease, and 1 patient is alive with progressive disease. There were no significant correlations between MYC mRNA expression levels and metastatic stage, age, or gender. A t test revealed that MYC mRNA expression was independent of TrkC mRNA expression (P = 0.60) and GFAP expression.

Univariate Cox regression analysis using MYC mRNA expression as a categorical variable showed that MYC mRNA expression was a significant prognostic factor for predicting progression-free survival (hazard ratio, 10.89; P = 0.024) and overall survival outcome (hazard ratio, 8.29; P = 0.047). When compared with the effect of metastatic stage, age, gender, extent of resection, therapy, TrkC expression, and GFAP expression, MYC mRNA expression was the most robust prognostic factor followed by TrkC mRNA expression (Table 3). Five-year overall and progression-free survival of the 26 patients are summarized in Table 4. Multivariable Cox regression analysis with inclusion of the clinical factors metastatic stage, age, and gender revealed that MYC mRNA expression remained a significant prognostic factor for progression-free survival (hazard ratio, 14.32; P = 0.013) and overall survival outcome (hazard ratio, 10.59; P = 0.034). The cumulative survival curves in the groups with high and low levels of MYC mRNA expression are shown in Fig. 2. The 5-year cumulative progression-free survival rate of the group with low levels of MYC mRNA expression was 92% and was significantly better than the 38% progression-free survival rate of the group with a high level of MYC mRNA expression (log-rank; P = 0.0048).

MYC gene amplification seemed not to be a predictor of tumor progression in the present study. Two of three patients (patients 3 and 26) whose PNET/MB showed evidence of MYC gene amplification have been followed for 70 months and are tumor free. However, the number of PNET/MB evaluated for MYC gene amplification was too small to include this variable in Cox regression models.

Information on TrkC mRNA expression was available for 19 of 26 PNET/MB patients. Despite the small sample size and limited number of clinical events, TrkC mRNA expression was a prognostic factor with a hazard ratio approaching statistical significance (hazard ratio, 3.78; P = 0.065). GFAP expression was not a prognostic factor in the present study. Whereas two of four patients (patients 6 and 12) whose PNET/MB was positive for GFAP were tumor free after 90 and 105 months of follow-up, two other patients (patients 21 and 23) had tumor progression. Univariate Cox regression analysis resulted in a nonsignificant hazard ratio of 1.23 (P = 0.76).

The combination of MYC mRNA expression and TrkC mRNA expression identified three risk groups of patients (Fig. 3). Children with PNET/MB expressing high TrkC mRNA and low MYC mRNA levels (n = 7) had the best survival outcome, with a progression-free survival rate of 100% after a median follow-up time of 55 months (range, 15–91 months). Of note, three of seven patients with good outcome survived without radiotherapy. Children with PNET/MB expressing high TrkC mRNA and high MYC mRNA levels (n = 7) had an intermediate progression-free survival outcome (71% at 5 years), whereas the group with low TrkC mRNA expression (n = 5) had the worst progression-free survival outcome (20% at 5 years) regardless of MYC mRNA expression. The prognostic model using MYC and TrkC mRNA expression was superior to known clinical factors in identifying a good outcome group of patients. Of 10 “standard-risk” patients (age at diagnosis ≥ 3 years, no evidence of metastatic disease, and ≥90% tumor resection), 5 (50%) died because of progressive disease.

The MYC proto-oncogene, a member of a family of highly related genes that includes MYCN and MYCL, encodes a nuclear phosphoprotein involved in the transcription of genes central to regulating the cell cycle (51), cellular proliferation (52), apoptosis (53), and embryonic development and differentiation (54). MYC proteins form heterodimers with Max, which is also a helix-loop-helix leucine zipper protein. MYC/Max heterodimers bind to specific DNA sequences located in the transcriptional control region of target genes and alter the transcription of these target genes by transactivation or transrepression.

MYC expression is normally tightly regulated throughout the cell cycle but may become deregulated or activated, contributing to malignant transformation (55). Dysregulation of MYC has been implicated in the pathogenesis of a variety of human neoplasms, including Burkitt’s lymphoma (reviewed in Ref. 56), breast carcinoma (57), colon adenocarcinoma (58), lung carcinoma (59), and prostate carcinoma (60). One important mechanism of MYC dysregulation results from gene amplification. MYC gene amplification has been identified in various cancers, including lung carcinoma (61) and breast carcinoma (62), and the presence of MYC gene amplification has been reported to correlate with unfavorable prognosis in breast carcinoma (63, 64). In childhood NB, amplification of MYCN, a MYC family member, occurs in about 25% of primary tumors and is strongly associated with advanced-stage disease, rapid tumor progression, and poor prognosis (65, 66, 67).

We found that there was a wide range of MYC mRNA expression and that MYC mRNA expression did not correlate with the presence of MYC gene amplification in PNET cell lines or primary PNET/MB. This finding is consistent with that of Bruggers et al. (50), who reported similar MYC mRNA expression levels in D425 cells (100-fold MYC amplification) compared with D283 cells (3-fold MYC amplification) and higher MYC mRNA expression levels in PFSK cells (single-copy MYC) compared with DAOY cells (single-copy MYC). Disparity between MYC gene copy number and MYC mRNA expression levels indicates the presence of alternative mechanisms to gene amplification in activating MYC and has been also reported for other cancers, including breast and colorectal carcinomas (58, 68).

In the present study, high levels of MYC mRNA expression correlated strongly with an unfavorable survival outcome in PNET/MB patients. MYC mRNA expression was more predictive than standard clinical factors that did not reach statistical significance on univariate analysis because of the relatively small samples size (26) and the limited number of clinical events (11). MYC mRNA expression levels did not correlate with metastatic stage, age, or gender of PNET patients. In multivariable analysis, correcting for these factors, the hazard ratio for MYC mRNA expression remained significant, indicating that MYC mRNA is of independent prognostic significance. Accordingly, MYC mRNA expression represents one of four individual biological prognostic factors we have identified in PNETs (20, 22, 25). Of these, TrkC mRNA and MYC mRNA expression seem to be the most potent prognostic factors when hazard ratios are compared and when the results of independent research groups are considered. Kim et al. (24) determined TrkC mRNA expression in 42 PNET/MB patients by Northern blot analysis and found that tumors with high TrkC mRNA expression were less likely to progress than tumors with low TrkC mRNA expression. Herms et al. (31) determined MYC expression in 72 PNET/MB by in situ hybridization, and they found that 30 patients with MYC mRNA-expressing PNETs/MBs had a worse survival outcome than did 42 patients whose tumors did not express MYC mRNA using their assay. In the present study, MYC mRNA expression was independent of TrkC mRNA expression and GFAP expression, suggesting that further analysis of these expression patterns be performed.

It is well known that the mechanisms regulating tumorigenesis are multifactorial. Therefore, it is unlikely that any single biological prognostic factor will be sufficiently robust to optimally segregate all PNET patients into risk-based strata. Accordingly, multiple factors of biological relevance are more likely to provide a more accurate system of defining risk groups. In the present study, we demonstrate this by combining MYC and TrkC mRNA expression. The group with low MYC and high TrkC mRNA expression had a 100% five-year progression-free survival outcome after a median follow-up time of 55 months (range, 15–91 months). Of note, three patients in this group were infants. Although the number of patients is too small to draw firm conclusions, it is conceivable that in this age group, where neurotoxic consequences of radiation therapy are greatest, a good outcome group can be identified by biological factors and potentially treated and cured with chemotherapy alone.

CNS PNETs and NB share clinical and biological similarities. They are both embryonal tumors derived from primitive neuroectodermal cells. They both occur predominantly in early childhood and represent the most common extracranial and intracranial neoplasms, respectively, in this age group. We now show that the combination of MYC mRNA expression and TrkC mRNA expression may be used in a prognostic model of CNS PNET/MB similar to the combination of MYCN gene amplification and TrkA mRNA expression in NB. Combination of TrkA mRNA expression and MYCN amplification identifies a low-risk group of NB patients (high TrkA expression and normal MYCN), an intermediate-risk group of NB patients (low TrkA expression and normal MYCN), and a high-risk group of NB patients (low TrkA expression and amplified MYCN; Ref. 69). Other well-established prognostic markers in NB are histological characteristics (70), deletions of chromosome 1p (71, 72), and gains of chromosome 17q (73). The most appropriate therapy for NB patients is determined not only by the use of clinical factors but also through the use of prognostication models including biological features (74, 75).

We anticipate that biological prognostic factors, in addition to clinical factors, will similarly define risk groups and help direct therapy decisions for children with PNET/MB. We therefore strongly recommend prospectively assessing MYC and TrkC mRNA expression levels in future PNET/MB clinical trials in an effort to validate this model for future incorporation in risk classification systems for clinical use.