Early Detection and Diagnosis
Outcome prediction and risk assessment by quantitative pyrosequencing methylation analysis of the SFN gene in advanced stage, high-risk, neuroblastic tumor patients
Article first published online: 22 JUL 2009
Copyright © 2009 UICC
International Journal of Cancer
Volume 126, Issue 3, pages 656–668, 1 February 2010
How to Cite
Banelli, B., Bonassi, S., Casciano, I., Mazzocco, K., Di Vinci, A., Scaruffi, P., Brigati, C., Allemanni, G., Borzì, L., Tonini, G. P. and Romani, M. (2010), Outcome prediction and risk assessment by quantitative pyrosequencing methylation analysis of the SFN gene in advanced stage, high-risk, neuroblastic tumor patients. Int. J. Cancer, 126: 656–668. doi: 10.1002/ijc.24768
- Issue published online: 8 DEC 2009
- Article first published online: 22 JUL 2009
- Accepted manuscript online: 22 JUL 2009 12:00AM EST
- Manuscript Accepted: 7 JUL 2009
- Manuscript Received: 2 MAR 2009
- Fondazione Italiana per la Lotta al Neuroblastoma
- Associazione Italiana per la Ricerca sul Cancro (AIRC)
- Regione Liguria (Progetto Diagnostica Avanzata)
- Italian Ministry of Health and Ministero dell'Università e della Ricerca Scientifica e Tecnologica (MURST), Italy
- DNA methylation;
- predictive oncology;
- Top of page
- Material and methods
- Supporting Information
The aim of our study was to identify threshold levels of DNA methylation predictive of the outcome to better define the risk group of stage 4 neuroblastic tumor patients. Quantitative pyrosequencing analysis was applied to a training set of 50 stage 4, high risk patients and to a validation cohort of 72 consecutive patients. Stage 4 patients at lower risk and ganglioneuroma patients were included as control groups. Predictive thresholds of methylation were identified by ROC curve analysis. The prognostic end points of the study were the overall and progression-free survival at 60 months. Data were analyzed with the Cox proportional hazard model. In a multivariate model the methylation threshold identified for the SFN gene (14.3.3σ) distinguished the patients presenting favorable outcome from those with progressing disease, independently from all known predictors (Training set: Overall Survival HR 8.53, p = 0.001; Validation set: HR 4.07, p = 0.008). The level of methylation in the tumors of high-risk patients surviving more than 60 months was comparable to that of tumors derived from lower risk patients and to that of benign ganglioneuroma. Methylation above the threshold level was associated with reduced SFN expression in comparison with samples below the threshold. Quantitative methylation is a promising tool to predict survival in neuroblastic tumor patients. Our results lead to the hypothesis that a subset of patients considered at high risk—but displaying low levels of methylation—could be assigned at a lower risk group.
Neuroblastic tumors (NTs) are the second most common childhood neoplasia and include two malignant histotypes, neuroblastoma (NB) and ganglioneuroblastoma (GNB), and the benign ganglioneuroma (GN).1
In the last years several clinical and biological criteria have been identified and are utilized to classify NTs patients into four “risk groups” (high, intermediate, low and very low) and this stratification has a direct impact on the clinical management of the patients and on the choice of the treatment regimen.2–6
Approximately 50% of the children with malignant NTs have metastatic disease at diagnosis, a condition that according to the International Neuroblastoma Staging System (INSS)5 classifies the patients as stage 4. The recent consensus approach for risk stratification established by the International Neuroblastoma Risk Group (INRG) task force,6 in agreement with other previous results,2–5 identified the INSS stage as the most significant prognostic variable and stage 4 patients are considered as a distinct group. Importantly, the distinction between stage 4 and stage 1, 2, 3 and 4S patients in the new INRG classification is at the top of the clinical decisional tree. Stage 4 patients are subdivided into three risk categories (high, intermediate and low) depending from their age at diagnosis, and, for the children younger than 18 months, from the MYCN amplification status and ploidy.2, 3, 6
Intermediate and low risk stage 4 patients generally perform well and their life expectancy at 60 months is higher than 70–80%.6–8 On the contrary, the majority of the high risk, stage 4 patients experience rapid and fatal disease progression and only 20–30% of them show progression free and overall survival longer than 60 months despite the introduction of novel multimodal therapeutic protocols. The favorable outcome observed in a subgroup of patients at high risk indicates that additional prognostic markers must be identified and validated to assign the patients to the most appropriate risk category and, ideally, to identify the most effective and less toxic treatment plan.4, 9, 10
In this respect, the aberrant hyper- or hypomethylation of gene promoter regions is considered a promising biomarker of outcome or response to treatment, and the potential clinical application of DNA methylation analysis is actively investigated.11, 12
In neuroblastic tumors, methylation profiles have been associated with well known predictors of clinical outcome13–18 and the observed hypermethylation of multiple CpG islands suggested the existence, in analogy with colorectal cancer,19 of a “CpG island methylator phenotype” (CIMP) associated with MYCN amplification and disease progression.15–18 These studies included highly heterogeneous cohorts of patients at various stages and assigned at different risk categories. Therefore, the presence of multiple and independent risk factors made it difficult to determine the full potential of DNA methylation analysis for the patients' evaluation and prognostication.
Most of the methylation studies on tumor samples were conducted by qualitative techniques and the results were generally expressed as dichotomous data (methylated or unmethylated).20, 21 However, this representation does not take into account the continuous variable nature of this epigenetic modification that is revealed by quantitative sequence-based techniques.22 Furthermore, the Methylation Specific PCR (MSP) analysis of tumor samples, even from microdissected material, often shows the concomitant presence of methylated and unmethylated alleles, whose presence cannot be attributed merely to infiltrating non cancer cells, confirming the heterogeneity of methylation in tumors.23
The significance of the “partial methylation” status observed in many studies has been generally overlooked. Quantitative changes of the methylation levels present individual variations in normal tissues,24 and, most importantly, in cancer tissues and during tumor progression25, 26 suggesting that the simple information of the presence or absence of methylated alleles may be inadequate to disclose important clinical and pathological features.
Along this line, the aim of our study was to establish if the quantitative analysis of DNA methylation, could identify threshold levels discriminating within the stage 4 high risk patients, those likely to have a favorable outcome from those presenting rapidly progressing and fatal disease and, in turn, if quantitative methylation analysis could be utilized to better define the risk group of these patients. This study was conducted utilizing a “class discrimination” strategy to determine the predictive value on the outcome at 60 months of different levels of methylation in a clinically homogeneous group of high-risk patients at stage 4.
Material and methods
- Top of page
- Material and methods
- Supporting Information
The Ethics Committee of the Giannina Gaslini Children Hospital of Genoa approved the collection, the storage in the Neuroblastoma Tissue Bank and the utilization of this material. Informed consent was obtained for all patients.
Genes selection, patients and planning of the study
The genes included in this study were selected on the basis of previous independent studies conducted on cohorts of patients at all stages and assigned at different risk groups, demonstrating that their hypermethylation could be a novel candidate prognostic marker.15, 16
SFN acts as a cancer suppressor by inhibiting the cell-cycle progression,27RASSF1A is a RAS effector that mediates cell proliferation and apoptosis,28CYP26C1 is a CYP26 family member involved in the metabolism of retinoids29 and DCR2 is a decoy member of the TRAIL receptor family.30
The clinical endpoints examined were the Overall Survival (OS) at 60 months and the Progression Free Survival (PFS) in relation with the level of methylation of the genes examined.
As the INSS stage is the major predictor of outcome in malignant NT patients,2, 6 to minimize the effect of this prognostic factor we determined the impact of the level of methylation on OS and PFS only in high risk patients at stage 4, that is the most common and aggressive mode of presentation of this disease.
The schematic representation of the study groups is reported in the Supporting Information Figure 1. The risk stratification criteria utilized were those of the International Neuroblastoma Risk Group (INRG) Classification System.6 Accordingly, stage 4 patients were considered at “high risk” if older than 18 months, irrespectively of other clinical and biological characteristics or, if younger than 18 months, when their tumor presented MYCN amplification. Patients below 18 months with MYCN-single copy tumors, were collectively considered as a control group at low or intermediate risk depending from the DNA index of the tumor.6
The study was initially conducted on a training cohort of 50 high-risk patients that were diagnosed between 1990 and 2003 with neuroblastoma (N = 37; range OS: 1–140 months) or ganglioneuroblastoma (N = 6; range OS: 14–164 months) or with malignant NT without further histological classification (N = 7; range OS: 7–72 months), and referred to the Giannina Gaslini Children Hospital, Genova, Italy (Table 1). Criteria of inclusion of these patients were that they should have presented progressive and fatal disease within 60 months from the diagnosis (“short survivors”) or they should have had a minimum follow up time of 60 months and should be alive at the last control (“long survivors”). Within this first set, 37 patients were included in the “short survivors” group (median OS: 28 months, range 1–56 months), and 13 were considered as “long survivors” (median OS: 89 months, range 61–165 months). MYCN amplification was not detected in the tumors of these latter patients. Ten long surviving patients were free of disease and 3 had stable disease at the last control (follow up time of 111, 126 and 131 months, respectively).
As control groups we included 16 stage 4 patients at intermediate and low risk. This group was composed by 13 neuroblastoma (range OS: 60–149 months), 2 ganglioneuroblastoma (OS: 87 and 111 months) and one NT tumor without further histological classification (OS 72 months). Ten of these patients were at low risk (DNA index > 1, range OS: 60–149 months); 4 were at intermediate risk (DNA index = 1, range OS: 87–129 months). DNA index was not determined in 2 patients (OS: 63 and 76 months). All these patients were alive and disease-free after a median follow up time of 89.6 months. As additional control group we included 10 patients with benign ganglioneuroma. These patients were diagnosed between 1994 and 2001, their median age at diagnosis was 92.8 months (SD: 41.5) and were all alive and free of disease at the last follow up (median follow up: 62.5 months, range: 14–134 months). None of these patients presented MYCN amplification or chromosome 1p alterations, DNA index was 1 in all samples.
We validated the methylation threshold risk-prediction model on an independent cohort of 72 consecutive patients, diagnosed between 1992 and 2004. This set of patients included 65 neuroblastoma (range OS: 5–151 months), 3 ganglioneuroblastoma (range OS: 19–66 months) and 4 malignant NT (range OS: 26–105 months) with no further histological classification (Table 1). The size of the validation set was defined according to a calculation based on 90% probability to detect a treatment difference at a two-sided 5.0% significance level if the true hazard ratio is 3.0. According to this calculation a minimum number of 47 patients was required. All the patients of the validation set were at stage 4 and at high risk; 55 of them were “short survivors” (median OS: 30 months) and 17 were “long survivors” (median OS: 105 months). However, differently from the training set, the validation cohort included 3 patients that died of disease 66, 68 and 89 months after diagnosis and that were thus classified as “long survivors”. Amplification of the MYCN oncogene was present in 5 of the long surviving patients of the validation set.
Patients that were alive but had a follow up time shorter than 60 months were excluded from the study.
Tumor samples and DNA isolation
The tumor DNA was retrieved from the Italian Neuroblastoma Tissue Bank.31 A pathologist examined the tumor tissue utilized for nucleic acid extraction to verify the identity and homogeneity of the samples. The specimens were collected at the onset of the disease, before therapy, and tumor cell content was at least 80%.
DNA was isolated by proteinase K digestion and phenol/phenol-chloroform extraction.
DNA Methylation was determined by pyrosequencing,32 a sequence-by-synthesis technique that allows the quantitative determination of the level of methylation of each of the CpG doublets within a target sequence.
DNA (1 μg) was modified by sodium bisulfite treatment as described by Frommer et al.22 and was subjected to pyrosequencing analysis. In a subset of patients, MSP21 and pyrosequencing were conducted in parallel. Experimental conditions and primers for MSP were previously described.15, 16 To verify the reproducibility of the bisulfite reaction, DNA derived from at least two different modifications was analyzed in the same MSP and pyrosequencing reaction.
For pyrosequencing analysis the modified DNA was first amplified by PCR to generate amplicons that included CpG sites in the regions previously studied by MSP by us and other groups.14–17 Blank reactions (for PCR and pyrosequencing) were included in each assay to exclude cross contamination. The amplicons were subjected to pyrosequencing in a Biotage PSQ 96MA system (Biotage, Uppsala, SW) utilizing the sets of reagents suggested by the supplier of the instrument. Primers were designed with the Pyrosequencing Assay Design software (Biotage, Uppsala, SW).
The specificity of the primers was determined in PCR reactions conducted on unmodified DNA to ensure that only the modified DNA was amplified.
The comparison between different pyrosequencing runs was conducted by inserting in random positions of the plates, control samples whose level of methylation was previously determined.
The sequence of the primers utilized for pyrosequencing, the target sequences and the PCR conditions are reported in the Supporting Information Table I. Each target sequence included 4 to 7 CpG doublets. MSP primers and reaction conditions were previously described.16
To determine the effect of DNA methylation on the expression of the SFN gene in neuroblastoma, we cultivated the LAN1 cell line in the presence of 5 μM 5-Aza-dC (MP Biomedicals, LLC. Heidelberg, GER) for 3 days. DNA was isolated from the cell by proteinase K digestion and phenol/phenol-chloroform extraction. Total RNA was extracted with Tryzol (Invitrogen, San Donato Milanese, IT). The level of methylation of SFN was evaluated by MSP and pyrosequencing. Induction of SFN expression by demethylation was determined by qPCR as described by Akahira et al.33 following the MIQE guidelines34 (Supporting Information Table II).
The expression of SFN in primary neuroblastic tumors was determined in 18 samples belonging to the training and to the validation cohort by qPCR according to the MIQE guidelines34 as described in the Supporting Information Tables II and III.
Because of the limited amount of available material, the tumor RNA was retrotranscribed and amplified utilizing the WT-Ovation™ RNA Amplification System kit (NuGEN Technologies, San Carlos, CA) according to the Manufacturer's instructions. Retrotranscription and amplification was conducted in parallel with positive and negative control RNA (total human reference RNA and LAN1 RNA, respectively).
The mean methylation value of the CpG doublets included in the target sequence was measured for each gene in all subjects, and this value was considered for the statistical analysis
All univariate comparisons between study groups were performed by ANOVA using the pairwise multiple comparisons test. An adjusted significance level of p < 0.05 was used to test hypotheses after Bonferroni correction. All significance tests were two tailed and CI was 95% in all cases.
Methylation values, as measured by pyrosequencing, were dichotomized according to the Receiver Operating Characteristic (ROC) curves, which identified those values providing the best separation between long and short survivors.35 The threshold values identified with this procedure (85% for SFN, 65% for DCR2, 60% for RASSF1A and 50% for CYP26C1) were utilized in all subsequent analyses.
Overall Survival is defined as the time elapsed from diagnosis to death. Only cancer-related deaths were considered. Patients that survived were censored at the last date they were reported to be alive. Progression Free Survival is calculated from the day of diagnosis to the date of relapse, as reported in the clinical records. Survival curves were computed according to the Kaplan-Meier method36 and were compared by means of the log-rank test.
The Cox proportional-hazards regression model was used to study in a multivariate setting the effect of DNA methylation on the overall and progression free survival. The following actual confounders: treatment protocol (Nb92, Nb97, other), clinical response (partial response, complete response, other), histology (neuroblastoma or ganglioneuroblastoma), and MYCN amplification (yes; no), age at diagnosis were included in all models. Although serum ferritin level (<92 or >92 ng/ml) does not provide clinically relevant information, it is of prognostic value in high risk, stage 4 patients and was included in the model.6
The association between technical repeats of methylation measures was estimated with the Pearson's correlation coefficient. Statistical analysis was carried out with SPSS for Windows Version 14.0 (SPSS, Chicago, IL), and STATA statistical software release 8 (StataCorp, College Station, TX).
Statistical analysis for expression data was performed with the non parametric Mann-Whitney test (two tailed).
- Top of page
- Material and methods
- Supporting Information
Comparison between MSP and pyrosequencing results and reproducibility of the assay system
In preliminary experiments we performed MSP and pyrosequencing analysis in a subset of 18 high-risk NT patients. In Figure 1 and in the Supporting Information Figure 2 we report three representative MSP and the Pyrosequencing results obtained for each of the four genes considered in our study. In 57 out of 72 determinations (79%), MSP analysis revealed the concomitant presence of methylated and unmethylated alleles independently from the patients' outcome. In these samples, the quantitative pyrosequencing showed methylation levels ranging from 25.6 to 90.3% for the four genes confirming the “partial methylation” status that was assigned by MSP.
The different intensity of the DNA amplification bands obtained by MSP in the reactions for the methylated and for the unmethylated alleles only partially reflects the absolute levels of methylation and likely depends also from the different efficiency of the two PCR reactions.
Pyrosequencing analysis showed that the CpG doublets, within the selected target sequences, had a homogeneous level of methylation (Supporting Information Fig. 2). To assess the reproducibility of the pyrosequencing determinations we compared the methylation levels of the SFN and RASSF1A genes in independent experiments and we observed a strong correlation (>95%) among the different repetitions (Supporting Information Fig. 3).
Given the consistency of this approach, and the absence of hot spots of hypermethylation in the CpG doublets considered, we used for the statistical analysis of each gene the mean value of methylation.
Quantitative methylation analysis in neuroblastoma patients
The comparison of the methylation status between short and long survivors, in the training cohort of patients at high risk, showed significantly higher levels of methylation for all genes in the patients that died of disease within 60 months (Table 2). We did not observe statistically significant differences in the level of methylation between the high-risk long survivors and the intermediate and low risk patients for the RASSF1A, CYP26C1 and DCR2 genes (p values comprised between 0.398 and 0.210), whereas a significant difference between these two groups of patients was observed for SFN (p = 0.039) (Table 2).
The predictive value of DNA methylation levels on the patients' outcome was analyzed in univariate and multivariate models utilizing the thresholds defined by the ROC curves (85% for SFN, 65% for DCR2, 60% for RASSF1A and 50% for CYP26C1).
As shown in Figure 2, Panel A, the Kaplan-Meier curves for OS and PFS estimates showed that the methylation levels of SFN, RASSF1A and CYP26C1 above the thresholds were strongly associated to a poorer outcome in the group of high risk patients (p values for the long-rank tests for OS and PFS were <0.001 for the three genes). Also the methylation of DCR2 above the threshold level was associated with reduced OS and PFS although at a lower statistical significance (p < 0.01).
The Cox regression analysis (Table 3), after the adjustment for the effect of know predictors of outcome (age at diagnosis, MYCN amplification, treatment protocol, ferritin, histology and clinical response), confirmed the strong predictive value on OS and PFS of methylation above the threshold for SFN and RASSF1A but not for CY26C1 and DCR2. The predictive value of hypermethylation of CYP26C1, observed in univariate analysis, was not confirmed in the multivariate model likely because the strong association between MYCN amplification and high levels of methylation of this gene (data not shown) made impossible to determine the independent contribution of these two markers to poor outcome.
Validation of the threshold levels of methylation
To validate the information provided by the training set of patients and to assess the robustness of our approach, we performed a similar analysis in an independent, cohort of 72 consecutive patients at stage 4.
Clinical and demographic variables were similar in case patients, both in the training and in the validation set, and no statistically significant differences in OS and PFS were observed between the two groups (OS p = 0.722; PFS p = 0.835) (Supporting Information Fig. 4). All the patients were diagnosed between 1990 and 2004 and the treatment protocols administered were comparable in the training and validation sets.
The major differences between the two cohorts were that the validation set included also patients that, although could be defined as “long survivors”, have died of disease progression after 60 months and that the tumors of 5 long surviving patients in the validation set, and none in the training set, were MYCN amplified.
The analysis of the validation set was restricted to SFN and RASSF1A as these were the only genes whose methylation thresholds significantly differentiated long and short survivors in a multivariate model.
As shown in Table 2, the difference between the mean of the methylation in the long and short survivors groups remained significant only for SFN. Similarly, the Kaplan Maier estimates for OS and PFS in the validation set and the multivariate analysis showed that SFN retained its predictive value as a biomarker of outcome (Fig. 2, Panel B, and Table 3).
The predictive significance of the methylation of RASSF1A above the threshold of 60% was not confirmed in the validation set. The stepwise analysis of the Cox regression model showed that several confounding factors, including MYCN amplification, contributed at a different extent to this result whereas no influence of any of the predictors was observed for SFN.
Furthermore, the quantitative methylation analysis of the SFN gene conducted on a set of benign ganglioneuroma showed that the level of methylation of these samples was not significantly different from that of the patients at intermediate and low risk and, most importantly, also from that of the long surviving patients at high risk from both the training and the validation cohorts (Table 4).
Distribution of DNA methylation in NT patients
Previous independent studies have shown that the level of methylation of some of the genes that define the CIMP phenotype in NT follows a bimodal distribution.15, 18 We have determined the distribution of DNA methylation in our casistic (combining the patients of the training and validation set for SFN and RASSF1A) and we have observed a trend toward bimodal distribution only for RASSF1A but not for SFN, CYP26C1 and DCR2 (Fig. 3).
Interestingly, the cut off methylation values discriminating between patients with a poor prognosis and long surviving patients determined with the distribution analysis, was essentially identical to that determined by ROC (data not shown).
Effects of DNA methylation on SFN expression in NT
Hypermethylation is a well-established mechanism for the control of gene transcription. The genes analyzed in this study are all down-modulated by methylation in a variety of tumors and methylation-dependent silencing of CYP26C1, DCR2 and RASSF1A, but not of SFN, has been already demonstrated in neuroblastic tumors.14, 17, 37
To determine if hypermethylation of the SFN promoter has a functional consequence also in NTs, we cultivated the neuroblastoma cell line LAN1, which bears a hypermethylated SFN gene (Fig. 4, Panel A), in the presence of the DNMT inhibitor 5-Aza-dC and we determined the level of expression of this gene by quantitative PCR analysis. The result of this experiment is reported in Figure 4, Panels A and B, and show that the reduction of the methylation level of the SFN promoter is accompanied by the16-fold increase of the expression level of the gene.
To determine if epigenetic mechanisms regulate SFN also in vivo, we have determined its level of expression in a subset of 18 NT mRNA samples derived from the training and from the validation cohorts. Within this subset, 11 samples had a methylation level above the discriminating threshold and the remaining 7 were below the threshold.
The relative quantification was performed utilizing ßactin as reference gene and as positive and negative controls, total human reference RNA and RNA from the neuroblastoma cell line LAN1, respectively. The result of this analysis, reported in Figure 4, Panel C, show that in 17/18 tumor samples SFN expression was absent or strongly reduced using as calibrator total human reference RNA. Overall, the expression of the samples whose methylation level was below 85% was significantly higher than that of the samples presenting a methylation level above the threshold (nonparametric Mann-Whitney test: p = 0.042). In 6 out of the 11 samples above the threshold, but only in 1 out of the 7 samples below the threshold, the level of expression of SFN was comparable or lower than that of the fully methylated LAN1 cell line. Overall this finding suggests the existence of a relation between the cut off value of methylation identified in the present study and the transcriptional activity of the SFN gene.
- Top of page
- Material and methods
- Supporting Information
The stratification of NT patients into “risk groups” is an essential step for the selection of the most appropriate treatment plan and to improve outcome.2–4, 6, 38 Nevertheless, even with the most aggressive multimodal therapies, the treatment of high risk, advanced stage NT patients remains a major challenge for the oncologist since only 20 to 30% of these patients have a life expectancy higher than 60 months.
In an attempt to define novel and stringent criteria for the more precise stratification of NT patients into risk groups and, in perspective, to help the design of optimized therapies, we have utilized the quantitative pyrosequencing methylation analysis technique, to detect epigenetic alterations predictive of outcome in high-risk patients at stage 4, that is the most common form of presentation of this disease.
In the last 5 years the interest for the utilization of epigenetic signatures for the molecular diagnostics and for the early detection of cancer has steadily increased.11, 12, 39, 40 In this respect, many qualitative or semiquantitative studies have shown that hypermethylation is a characteristic of the most aggressive NTs.13–17 However, the dichotomization of methylation data provided by the qualitative or semiquantitative techniques that were utilized in those studies and the heterogeneity of the patients enrolled, likely has masked major clinical differences. Indeed the MSP analysis in a subset of the homogeneous cohort of stage 4, high risk patients analyzed in our study, showed the concomitant presence of the methylated and of the unmethylated target in 79% of the samples independently from the patients' outcome. This finding suggests that the simple determination of the presence or absence of methylated targets might be a reductive approach that cannot disclose the full potential of epigenetic analysis. Indeed, methylation is a dynamic process and methylation levels may change during tumor progression,25, 26 suggesting that the quantitative analysis might identify novel and potentially highly informative prognostic biomarkers in NTs and other tumors.
Bisulfite sequencing22 of cloned PCR products is considered the “gold standard” for quantitative methylation analysis since it can generate the precise methylation map of the target region. This time-consuming procedure is now being superseded by qPCR-based procedures, like MethyLight,41 and by pyrosequencing,32 two faster techniques, still based on the bisulfite conversion of the unmethylated C into T. These novel methodologies can detect minimal amount of DNA methylation and, between the two, we chose pyrosequencing because it allows to directly measure the methylation level of each CpG in the target sequence and to exclude methylation heterogeneity in the analyzed region.
Our retrospective study was conducted initially on a clinically homogeneous training set of patients at stage 4 that were all classified at high risk and subdivided into “long” and “short survivors” according to their survival time (> 60 months or < 60 months, respectively). The results obtained with this set were then validated on an independent cohort of consecutive patients. Differently from other studies, pyrosequencing enabled us to consider methylation as a continuous variable and to apply a standard ROC curves approach to determine the values which best discriminate the patients according to their outcome.
The distribution of DNA methylation observed in our casistic confirmed, for RASSF1A, the trend toward a bimodal distribution already described in other studies for different genes15, 18 whereas bimodality was not as well defined or absent for the other genes. Considering the type of casistic analyzed in our study (only stage 4 patients at high risk), this finding was not surprising. Indeed, the studies where the bimodal distribution of methylation levels was described15, 18 included patients at all stages and high levels of methylation were prevalent in advanced stage patients like the casistic analyzed in this report.
The result of our study showed that the methylation of the SFN gene above a defined threshold is a strong and reliable predictor of adverse outcome independently from other prognostic factors. Interestingly, we have previously reported that the SFN gene is partially or fully methylated in benign and malignant neuroblastic tumors.16 However, in that early study, the limits of the qualitative analysis utilized did not allow to detect the predictive value of SFN methylation on outcome that was instead fully disclosed by the quantitative determination herein reported.
The strong predictive value on outcome of RASSF1A methylation above the threshold level of 60% that was observed in the training cohort was not confirmed in the validation set because of the contribution of the other predictors of adverse outcome in the unselected patients' population.
The “methylator phenotype,” determined by quantification of the methylation level of multiple genes,15 was recently validated as a significant predictor of PFS in neuroblastoma patients.18 One of the questions that remain to be answered is how the prognostic power of SFN methylation compares to that of CIMP. In this respect we have confirmed that the level of methylation of CYP26C1, one of the genes included in the original set that defined the CIMP phenotype, is significantly higher in short surviving as compared to long surviving patients. However, this gene was excluded from the validation analysis since, in multivariate analysis, its predictive value was lost likely because the strong association between MYCN amplification and high levels of CYP26C1 methylation made impossible to determine the independent contribution of these two markers to poor outcome (Ref.15 and our data not shown).
The quantitative PCR analysis conducted on a subset of patients demonstrated that the overall expression of SFN is significantly lower in the samples with a methylation level above 85% as compared to that of the tumors presenting lower methylation. This finding indicates that the methylation threshold identified in our study has a direct functional consequence and suggests that SFN is a novel candidate gene involved in neuroblastic tumors.
SFN belongs to the evolutionary highly conserved 14.3.3 gene family that participates to many crucial functions and pathways like the maintenance of the G2 cell cycle checkpoints, DNA repair, apoptosis, cellular senescence and cell adhesion and motility.27SFN is a downstream effector of p53 activated in response to DNA damage and has been implicated in many epithelial tumors.27 Although SFN expression traditionally was considered restricted to epithelial cells, the qPCR analysis herein reported in addition to other our unpublished observations and to data retrieved from expression microarray databases (http://www.ncbi.nlm.nih.gov/geo/), showed that SFN could be detected in a variety of normal and tumor tissues of different species and in a subset of neuroblastoma cell lines and tumors.
Although additional work will be obviously necessary to determine if SFN has a specific role in NT pathogenesis, it is of interest to observe that the overexpression of SFN inhibits AKT, an oncogene involved in cell survival and a downstream target of TRK, a gene implicated in NT pathogenesis and in its clinical behavior.42, 43
Furthermore, it has been demonstrated that the p53 homologue p73 is a more efficient activator of SNF than p53 itself44 and that the transactivation-deficient variant of the third member of the p53 family, ΔNp63, acts as a strong repressor of SNF transcription.45 It will be of interest to determine if also the transactivation-deficient and oncogenic p73 variant ΔNp73, that when overexpressed is a molecular marker of adverse outcome in NT,46 acts as a repressor of SNF transcription.
Neuroblastoma and ganglioneuroblastoma are among the first tumors where biological factors like MYCN amplification, ploidy, and expression of selected genes or chromosomal rearrangements and deletions, were recognized as critical determinants of the patients' outcome38 and some of them are utilized to define risk categories and to optimize therapeutic strategies.2–4, 6 In this respect, our results lead us to hypothesize that low levels of methylation of the SFN gene could identify a subset of NT patients that by standard criteria are considered at high risk but that instead could be assigned at a lower risk group either because their disease is less aggressive or is more responsive to treatment. It can be foreseen that microarray techniques, capable to detect multiple targets of aberrant methylation, could be coupled with quantitative methylation analysis to develop new generation platforms for a more accurate prediction of the patients' outcome and, ideally for their assignment to the most appropriate risk category and to the optimal treatment plan.
In conclusion, our data demonstrate the power of quantitative DNA methylation analysis in risk assessment and to the best of our knowledge, define for the first time a threshold level of methylation associated with the clinical characteristics and survival of advanced stage, high risk NT patients.
Furthermore, the observation that survival is tightly linked to the extent of methylation provides a strong rationale to the exploitation of the epigenome as a target for innovative experimental therapies in neuroblastic tumors.
- Top of page
- Material and methods
- Supporting Information
The authors thank the patients and their families for the constant co-operation. They also thank the physicians of the Giannina Gaslini Children Hospital, Genova, that have provided clinical records and Prof. Riccardo Rosso (IST-Genova) and Dr. Bruno De Bernardi (G. Gaslini Children Hospital, Genova) for their suggestions, advice and support. Dr. Barbara Banelli and Dr. Katia Mazzocco are fellows of the Fondazione Italiana Per la Lotta al Neuroblastoma.
- Top of page
- Material and methods
- Supporting Information
- 1Terminology and morphologic criteria of neuroblastic tumors: recommendations by the International Neuroblastoma Pathology Committee. Cancer 1999; 86: 349–63., , , , , .
- 2Neuroblastoma: biological insights into a clinical enigma. Nat Rev Cancer 2003; 3: 203–16..
- 3The International Neuroblastoma Risk Groups (INRG): a preliminary report. Eur J Cancer 1997; 33: 2113–6., , , , , , , , , , , , et al.
- 4Neuroblastoma. Lancet 2007; 369: 2106–20., , , .
- 5Revisions of the international criteria for neuroblastoma diagnosis, staging, and response to treatment. J Clin Oncol 1993; 11: 1466–77., , , , , , , , , .
- 6The International Neuroblastoma Risk Group (INRG) classification system: an INRG Task Force report. J Clin Oncol 2009; 27: 289–97., , , , , , , , , , , , et al.
- 7Evidence for an age cutoff greater than 365 days for neuroblastoma risk group stratification in the Children's Oncology Group. J Clin Oncol 2005; 23: 6459–65., , , , , , , , , , .
- 8The role of age in neuroblastoma risk stratification: the German. Italian, and children's oncology group perspectives. Cancer Lett 2005; 228: 257–66., , , , , , , , , .
- 9Evolving significance of prognostic markers associated with new treatment strategies in neuroblastoma. Cancer Lett 2003; 197: 119–24., , .
- 10The biologic basis for neuroblastoma heterogeneity and risk stratification. Curr Opin Pediatr 2005; 17: 7–13..
- 11The power and the promise of DNA methylation markers. Nat Rev Cancer 2003; 3: 253–66..
- 12Detection and interpretation of altered methylation patterns in cancer cells. Nat Rev Cancer 2005; 5: 223–31..
- 13Clustering of gene hypermethylation associated with clinical risk groups in neuroblastoma. J Natl Cancer Inst 2004; 96: 1208–19., , , , .
- 14Association of epigenetic inactivation of RASSF1A with poor outcome in human neuroblastoma. Clin Cancer Res 2004; 10: 8493–500., , , , , , , , .
- 15CpG island methylator phenotype is a strong determinant of poor prognosis in neuroblastomas. Cancer Res 2005; 65: 828–34., , , , , , , , .
- 16Distinct CpG methylation profiles characterize different clinical groups of neuroblastic tumors. Oncogene 2005; 24: 5619–28., , , , , , , , .
- 17Methylation of CASP8. DCR2, and. HIN-1 in neuroblastoma is associated with poor outcome. Clin Cancer Res 2007; 13: 3191–7., , , , , , , .
- 18Marked and independent prognostic significance of the CpG island methylator phenotype in neuroblastomas. Cancer Lett 2007; 247: 253–8., , , , , .
- 19CpG island methylator phenotype in colorectal cancer. Proc Natl Acad Sci USA 1999; 96: 8681–6., , , , , .
- 20DNA methylation: a profile of methods and applications. Biotechniques 2002; 33: 632, 4, 6–49., .
- 21Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands. Proc Natl Acad Sci USA 1996; 93: 9821–6., , , , .
- 22A genomic sequencing protocol that yields a positive display of 5-methylcytosine residues in individual DNA strands. Proc Natl Acad Sci USA 1992; 89: 1827–31., , , , , , , .
- 23p16(INK4a) promoter methylation and protein expression in breast fibroadenoma and carcinoma. Int J Cancer 2005; 114: 414–21., , , , , , , , , .
- 24Epigenetic differences arise during the lifetime of monozygotic twins. Proc Natl Acad Sci USA 2005; 102: 10604–9., , , , , , , , , , , , et al.
- 25Massively parallel bisulphite pyrosequencing reveals the molecular complexity of breast cancer-associated cytosine-methylation patterns obtained from tissue and serum DNA. Genome Res 2008; 18: 19–29., , , , , , , , , , .
- 26Quantitative assessment of promoter hypermethylation during breast cancer development. Am J Pathol 2002; 160: 605–12., , , , , .
- 27The 14–3-3 cancer connection. Nat Rev Cancer 2003; 3: 931–43..
- 28Role of the Ras-association domain family 1 tumor suppressor gene in human cancers. Cancer Res 2005; 65: 3497–508., , .
- 29A novel human cytochrome P450. CYP26C1, involved in metabolism of 9-cis and all-trans isomers of retinoic acid. J Biol Chem 2004; 279: 77–85., , , , , , , .
- 30The novel receptor TRAIL-R4 induces NF-kappaB and protects against TRAIL-mediated apoptosis, yet retains an incomplete death domain. Immunity 1997; 7: 813–20., , , , , .
- 31Neuroblastoma: a multiple biological disease. Eur J Cancer 29A: 802–4, 1993..
- 32De novo quantitative bisulfite sequencing using the pyrosequencing technology. Anal Biochem 2004; 333: 119–27., , , .
- 33Decreased expression of 14–3-3 sigma is associated with advanced disease in human epithelial ovarian cancer: its correlation with aberrant DNA methylation. Clin Cancer Res 2004; 10: 2687–93., , , , , , , , , , , .
- 34The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem 2009; 55: 611–22., , , , , , , , , , , .
- 35A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983; 148: 839–43., .
- 36Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457–81., .
- 37Identification of genes targeted by CpG island methylator phenotype in neuroblastomas, and their possible integrative involvement in poor prognosis. Oncology 2008; 74: 50–60., , , , , , .
- 38Association of multiple copies of the N-myc oncogene with rapid progression of neuroblastomas. N Engl J Med 1985; 313: 1111–6., , , , , , .
- 39Cancer diagnosis, risk assessment and prediction of therapeutic response by means of DNA methylation markers. Dis Markers 2007; 23: 89–96., .
- 40Use of DNA methylation for cancer detection and molecular classification. J Biochem Mol Biol 2007; 40: 135–41., .
- 41Laird PW MethyLight: a high-throughput assay to measure DNA methylation. Nucleic Acids Res 2000; 28: E32., , , , , , ,
- 42DNA damage-induced protein 14–3-3 sigma inhibits protein kinase B/Akt activation and suppresses Akt-activated cancer. Cancer Res 2006; 66: 3096–105., , , , , , , , , .
- 43Biological effects of TrkA and TrkB receptor signaling in neuroblastoma. Cancer Lett 2005; 228: 143–53., , , , , , , , , .
- 44The potential tumor suppressor p73 differentially regulates cellular p53 target genes. Cancer Res 1998; 58: 5061–5., , , .
- 45The delta Np63 alpha phosphoprotein binds the p21 and 14–3-3 sigma promoters in vivo and has transcriptional repressor activity that is reduced by Hay-Wells syndrome-derived mutations. Mol Cell Biol 2003; 23: 2264–76., , , .
- 46Expression of deltaNp73 is a molecular marker for adverse outcome in neuroblastoma patients. Cell Death Differ 2002; 9: 246–51., , , , , , , , .
- Top of page
- Material and methods
- Supporting Information
Additional Supporting Information may be found in the online version of this article.
|IJC_24768_sm_suppfigure1.tif||3700K||Supporting Figure 1.|
|IJC_24768_sm_suppfigure2.doc||440K||Supporting Figure 2.|
|IJC_24768_sm_suppfigure3.tif||6593K||Supporting Figure 3.|
|IJC_24768_sm_suppfigure4.tif||72K||Supporting Figure 4.|
|IJC_24768_sm_supptable1.doc||44K||Supporting Table 1.|
|IJC_24768_sm_supptable2.doc||126K||Supporting Table 2.|
|IJC_24768_sm_supptable3.xls||15K||Supporting Table 3.|
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.