Background
Central nervous system (CNS) pediatric tumors of embryonal origin are a heterogeneous group of malignant neoplasms that comprise by far the largest group of malignant brain tumors in childhood. All embryonal tumors are classified pathologically as grade IV malignancies and they are largely comprised of medulloblastomas (MBs) and atypical teratoid/rhabdoid tumors (AT/RTs), among others [
1]. Medulloblastoma is considered the most common malignant pediatric brain tumor, accounting for 20% of cases [
2]. Risk stratification based on clinical parameters is insufficient for accurate prognostication [
3]. Current treatment options include surgery, chemotherapy and radiotherapy [
4]. However, age limitations for the delivery of radiotherapy have been set due to the vulnerability of the developing brain to radiotherapy-induced neurocognitive deficits [
5]. Notwithstanding, aggressive multimodal therapy has improved the prognosis for children with MB, nearly one third of patients will eventually succumb to progressive tumors. Atypical teratoid/rhabdoid tumour is also a highly malignant CNS tumour that represents 1-2% of pediatric brain tumours and accounts for at least 10% of CNS tumours in infants, due to the predominance in children younger of 3 years old. The survival of children younger than 3 years of age remains poor, particularly for patients with supratentorial tumors and those with metastatic disease [
2]. It can be supratentorial, especially in cerebral hem [
6] ispheres, or infratentorial, especially in the cerebellar hemispheres, cerellopontine angle and brain stem mainly in children younger than 2 years of age. Prognosis of AT/RT is dismal, while there are no protocols aimed specifically for this type [
2].
The mainstay treatment for both local and distant disease control in older children, normally >3 years remains craniospinal irradiation [
7]. Nevertheless, outcomes remain highly associated with increased mortality due to current therapy-resistant disease. It is therefore likely that improved treatments will only be possible when the tumor-specific molecular events are better understood.
Growing evidence suggests that a class of small cellular RNAs, termed microRNAs (miRNAs) play important roles in controlling the development of the CNS, by regulating cell proliferation, differentiation and apoptosis [
8],[
9]. Recent miRNA-based research reveals that a) miRNA profiles of tumor cells are dissimilar from normal cells and b) miRNA expression profiles in tumors from similar developmental origins have similar alterations [
10]. Understanding the interplay between normal brain development and CNS tumors pathogenesis is essential for the development and implementation of more efficient and less toxic targeted therapies [
4]. Given these, and the fact that miRNAs are less susceptible to chemical modification and RNase degradation, provide the rationale that miRNAs might afford substantial value for neuro-oncology. Thus far, approaches to miRNA-targeted therapies have been mentioned in lung cancer, lymphoma and pancreatic cancer [
11]-[
14]. Regarding brain tumors, to the best of our knowledge, there are no reports available of miRNA use for therapeutic purposes, except for
in vitro systems [
15],[
16]. In particular, the use of miR-34a in several cancer models is encouraging, since it has been found down-regulated in diverse types of tumors [
17]-[
22] and its expression might suppress the ontogenesis of those specific types of malignancies.
Undeniably, thus far, expression profile analysis has revealed several miRNA signatures related to malignant pediatric brain tumors [
9],[
23]-[
27]. However, as the number of novel miRNAs is still increasing, large-scale screening is necessary to profile the global miRNA expression. To our knowledge, only a few studies applying miRNA microarrays have been conducted [
24],[
28],[
29] in pediatric brain tumors. MiRNA microarrays afford the most commonly used tool for the large-scale screening of miRNA expression. Subsequently, bearing in mind that a) miRNAs represent one of the largest classes of gene regulators and b) the clinical significance of miRNA expression profiles as well as their biological role in pediatric CNS embryonal tumors remains to be elucidated, characterization of miRNA patterns in pediatric malignant MBs and AT/RTs might have substantial value for diagnostic and prognostic purposes as well as for advanced therapeutic interventions.
In the current setting we performed for the first time miRNA microarrays and meta-analyses based on miRNA microarrays datasets from GEO to identify and validate miRNA candidates relevant to pediatric CNS embryonal prognosis. The obtained miRNA patterns were correlated to several factors known to affect survival including tumor type, age at diagnosis and tumor recurrence. We found that certain miRNA profiles were relevant to each of the aforementioned variables and their possible combinations delivering potential reliable diagnostic, prognostic or therapy-related targets, while this is the first report that underlines the adverse role of miR-34a and its association with embryonal brain tumor patients’ inferior prognosis.
Discussion
Previous research in various cancer disease processes suggested that miRNAs play vital roles in disease pathogenesis and have potential as biomarkers and therapeutic agents. In particular, the use of miRNAs as embryonal tumor markers constitutes an exciting field in biomedical research, due to diverse aspects including ability to predict disease before the onset of clinical symptoms, rich information content, great discriminatory power, accessibility in different specimen types, possibility of being evaluated from different sources and their potential for accurate quantification following the application of high-throughput technologies [
30]. Yet, principally due to the fact that the CNS is the least accessible of all tissues and the lack of clinically related animal models for miRNA research, no embryonal tumor miRNA biomarkers have been established in clinical practice [
30],[
31].
In the current setting we performed miRNA microarrays in pediatric patients with embryonal CNS neoplasms to investigate whether the differential expression of miRNA genes is associated with factors that envisage the clinical course of the disease. Ultimately, we aimed to identify and provide additional data regarding potential miRNA expression signatures relative to prognosis of pediatric MBs and AT/RTs, specifically. At present, there have been comparatively limited reports focusing on identifying miRNA biomarkers related to MB and AT/RT diagnostics and therapeutics [
23],[
26],[
28],[
32]. To our knowledge, this is the first attempt aiding specifically to evaluate an association between miRNA expression patterns and disease progression or patient outcome in pediatric MBs and AT/RTs and this is the first attempt that used meta-analysis to validate findings.
By performing miRNA microarrays, we identified overall 113 differentially expressed miRNAs between embryonal tumors and the control cohort. Among them, 107 were tissue-specific and 6 miRNAs were common between the two tumor types. More specifically, miR-34a, miR-548j, miR-3191 and miR-3912 were found up-regulated in both MBs and AT/RTs when compared to the control group. Based on previous reports, miR-34a is considered a putative tumor suppressor gene in various types of cancers [
33]-[
38] including brain tumors [
39]-[
42]. MiR-34a regulates the expression of SIRT1 (silent information regulator 1), which functions as an oncogene by inactivating key tumor suppressor proteins such as the transcription factor p53 [
43]-[
45]. Its targets include factors required for cell cycle progression, anti-apoptotic proteins and proteins involved in invasion [
46].
In the current study, almost all analyses and associations confirmed the up-regulation of miR-34a in the embryonal tumors group and in the MB group alone. Overexpression of miR-34a has been previously observed in gastric cancer [
47] as well as in pediatric brain tumors including medulloblastomas [
23], ependymomas [
48] pilocytic astrocytomas [
28] and in low- and high-grade astrocytomas [
49]. More specifically, our meta-analyses using the qRT-PCR raw data from Ferretti et al. [
23], revealed that miR-34a was up-regulated in the 34 patients with MB that they screened. Additionally, according to Costa et al. miR-34a was found up-regulated in all 34 pediatric ependymomas that they tested with qRT-PCR against all normal samples. In Birks et al. [
28] report, it is also obvious that miR-34a was overexpressed in all four pilocytic astrocytomas that they compared against a total of eight normal pediatric brain specimens, using miRNA microarrays. In addition, our meta-analysis using their raw data revealed that miR-34a was also overexpressed in all four MBs that they screened with miRNA microarrays. Findings from Birks et al. [
28] are also in line with those reported by Ruiz Esparza-Garrido et al. [
49] who later demonstrated that miR-34a was up-regulated in all twelve pediatric astrocytomas (3 pilocytic astrocytomas, 4 diffuse astrocytomas and 5 glioblastoma multiforme astrocytomas) they screened and compared to four control brain tissues. Bearing in mind our miR-34a findings, which a) have been validated with both additional experimentation and meta-analyses using more miRNA microarrays datasets and b) are in line with all aforementioned previous reports, it seems likely that miR-34a might not play global tumor suppressor roles. For reliable conclusions to be drawn this should be validated in even larger patient cohorts.
Using a strict filtering approach in the ROC tests, we determined a group of miRNAs that were also validated in an independent patient cohort. Therefore, we suggest that these miRNAs could be successfully used as distinguishing markers for each tumor type or between tumor types and normal tissues. It is noteworthy, that among them, once again, miR-34a was observed. In addition, we showed that miR-548j, and miR-3912, which have already manifested differential expression between embryonal tumors and the control cohort, have a high potential to distinguish AT/RTs alone from normal tissues, affording significant potential diagnostic value for AT/RTs. The role of miR-548j and miR-3912 is described for first time in embryonal tumors, as no previous related reports were found. Moreover, we observed that miR-192 and miR-3617 could accurately discriminate MBs over the normal tissues indicating their potential as putative biomarkers for MB diagnostics.
Overexpression of miR-192 contributes to tumor growth and progression in pancreatic ductal adenocarcinoma [
50], and has been associated with early diagnosis of distant metastasis of gastric cancer [
51], and with clinical relevance and prognostic significance in colon cancer [
52]. Regarding brain tumors, our results are in line with Ruiz-Esparza-Garrido et al. [
49], who manifested that miR-192 was up-regulated in 5 high-grade astrocytomas, when compared to 7 low-grade ones. On the contrary, though, Liu et al. [
53] and Ferretti et al. [
23] suggested that miR-192 was found significantly down-regulated in eight pediatric gliomas and in 34 pediatric MBs, respectively, whilst other reports in pediatric brain tumors have not identified this gene as differentially expressed [
28],[
29],[
48],[
49],[
54]. Given these, the role of miR-192 in childhood CNS malignancies remains to be elucidated. No reports were found regarding the value of miR-3617 in similar study groups.
Hierarchical clustering analysis using the miRNA signatures revealed that AT/RTs formed distinct group versus the controls, therefore assisting in the diagnosis of this type of neoplasias. However, this was not demonstrated in MBs. We would agree with Sredni et al. [
32] that miRNA expression profiling might not afford the finest tool to classify brain tumors for diagnostic purposes, especially in studies with small patient cohorts. Additional studies have also failed to cluster absolutely correct (100%) brain tumors based on miRNA expression patterns [
24],[
25],[
28],[
55]. To validate this result, we have utilized an independent group of GE tumors. Germinomas are considered childhood tumors of embryonic origin, arising from primordial germ cells that have migrated aberrantly during embryonic development and subsequently undergone malignant transformation [
56]. To investigate whether the inability of HCL was due to samples’ relativeness or other technical issues, we analyzed all data (MBs, AT/RTs, GEs) as a common cohort. Subsequently, HCL successfully discriminated between GEs and MBs or AT/RTs. This finding might indicate that MBs and AT/RTs are very close both in origin and in expression profiles. To our knowledge, miRNA expression has been found indisputably distinct between tumor types or tumors versus controls in one previous occasion [
23].
The prognosis of early childhood medulloblastoma treated with surgery, radiotherapy and chemotherapy has been dismal (20%–45%) compared with older children, probably due to a diverse biology underlying medulloblastoma in younger children [
57]. In addition, whether the molecular alterations described in embryonal tumors of older children are consistent in infants remain to be confirmed [
7]. For this purpose we divided our patients to different age groups based on treatment intervention and we investigated potential distinct miRNA expression profiles. Overall, 11 miRNAs were found aberrantly expressed. Among them, the most imperative miRNAs with higher levels of expression obtained in youngest children (Group A) and lower levels of expression in older children (Group C) included miR-720 and miR-3681. It is noteworthy, that 62.5% (5/8) of children remain alive in Group A, although none of the children in this Group received radiotherapy. In Group C, 57.1% (4/7) of patients deceased from the embryonal tumor. In this Group, all children have received intense radiotherapy in combination with chemotherapy and surgery. It seems likely, that overexpression of miR-720 and miR-3681 confers an additional advantage to patients at higher-risk, whereas down-regulation of these miRNAs reflects dismal prognosis to lower-risk groups of patients. No reports were found about the value of miR-720 and miR-3681 in pediatric CNS malignancies.
In the current study we also endeavored to identify gender-specific miRNAs that could potentially contribute susceptibility to embryonal brain malignancies. According to Pinheiro et al. [
58] notwithstanding the value of the X-linked miRNAs remains to be elucidated, several miRNAs participate in cancer onset and progression, while male preponderance has been observed in diverse types of cancer. Nevertheless, among the differentially expressed miRNAs identified in this setting, none of them was X located.
In an attempt to ascertain which miRNA signatures are implicated in the progression and outcome of the disease, we performed correlations between the following groups: control tissue cohort and patient cohort including relapsed patients and patients in complete remission as well as control tissue cohort and patient cohort including deceased patients and patients that remain alive, respectively. Based on our findings, following both initial and meta-analyses, we report the up-regulation of miR-3681, miR-34a, miR-136, miR-26b and miR-192 in the patient group, subsequently leading to the conclusion that they might possess oncogenic activities. The expression profiles initially obtained for miR-320e and miR-642a were not validated with meta-analyses. In addition, it is noticeable that the overexpression of miR-34a that we detected in previous correlations was once more confirmed. This is in line with Costa et al
.[
48] who also found miR-34a highly overexpressed in ependymomas, however, according to their results miR-34a portends a more favorable prognosis, instead. Yet again, miR-192 was found up-regulated in the patient cohort, underlying the potential of this gene, which according to our findings it emerges not only as a putative diagnostic biomarker, but also as predictor for the likelihood of an inferior clinical outcome.
Moreover, from current associations, although initially miR-601 emerged as a putative oncogene, neither meta-analysis nor qRT-PCR confirmed this finding. Therefore, it seems likely that this gene has potential tumor suppressor properties and our initial miRNA microarrays analysis was inadequate to reveal this. Previous reports were found describing both oncogenic and tumor suppressor roles for miR-601. For instance, elevated expression levels for miR-601 have been previously reported in pediatric ependymomas versus the control cohort and in pediatric glioblastomas when compared to anaplastic astrocytomas [
25],[
48]. On the contrary, miR-601 might serve as a novel biomarker for the clinical diagnosis of colorectal cancer when down-regulated [
55],[
59]. Based on our results, we propose that miR-601 plays putative tumor suppressor roles in MBs and its up-regulation is subsequently associated with favorable prognosis. Also, it is probable that it exerts tissue specific roles.
Remaining in the same context, both initial and meta-analyses showed that miR-720, miR-891a, miR-3665, miR-3620, miR-382, miR-452, and miR-122 were down-regulated in the patient group and therefore indicating that they might possess tumor-suppressive activities. Previous reports also proposed a possible link between these tumor-suppressor miRNA genes with diverse types of malignancies, on several occasions. For instance, reduced levels of miR-720 and miR-382 have been hitherto demonstrated in breast cancer [
60],[
61] and MBs [
23], respectively. Down-regulation of miR-452 has been associated with promoting stem-like traits and tumorigenesis of gliomas and it has been suggested that it may represent a novel prognostic biomarker and therapeutic target for the disease [
62]. In addition, miR-122, has been well-documented to act as tumor-suppressor gene for hepatocellular carcinoma [
63],[
64] and breast cancer [
65]. Regarding brain tumors, under expression of miRNA-122 has been demonstrated in glioma specimens and glioma cell lines and has been inversely correlated with patients’ survival following surgery [
66].
It is noteworthy, that the vast majority of the aforementioned miRNA oncogenes and tumor-suppressor genes were also observed when comparisons were made between miRNA expression patterns and the disease outcome, which verifies and validates their putative predictive role, sometimes favorable and sometimes poor, in embryonal CNS malignancies. These results were also confirmed by the regression analysis, which manifested both differential expressions in those miRNAs as well as linear behavior between tumor types.
Functional annotation of the miRNA cohort indicated that the DE miRNAs participate in RNA polyadenylation, which includes RNA processing, autophagy and pyrimidine synthesis. To the best of our knowledge, there are no reports available regarding such associations. It could be hypothesized that the identified miRNA cohort plays a role in tumor progression, since autophagy has been reported to be a survival mechanism in tumor cells [
67]. At the same time, RNA processing and nucleotide synthesis are also signs of progressive disease, which probably could be the case for the present miRNA cohort. Interestingly, the identified miRNAs appeared to be of importance in tumors as signified by the
Disease Association annotation analysis performed.
In summary, the current study was shown to provide significant insights into the growing role of several miRNA signatures in pediatric embryonal CNS neoplasms including MBs and AT/RTs. Following associations with certain patients’ clinical characteristics and cross-validation with meta-analyses, there was good evidence that miR-601 emerged as a putative tumor suppressor gene in MBs, whereas miR-34a manifested potential oncogenic roles. This is the first study to report the adverse properties of miR-34a as demonstrated in pediatric embryonal brain neoplasms, indicating tissue-specific roles rather than global tumor suppressor properties. Collectively, our findings modulated novel molecular biomarkers which might have a promising potential in pediatric embryonal CNS malignancies.
Authors’ contributions
MB conceived and designed the study, performed all experiments, evaluated and interpreted data analyses, and drafted the manuscript. GIL performed all data analyses, participated in interpretation of data analyses and in drafting the manuscript, GK assisted in all experiments performance, VM assisted in immunohistochemistry performance, SK performed tumor diagnosis, DKB provided miRNA microarrays raw data for meta-analyses, NP performed all tumor resections, AKol assisted in miRNA microarray performance, AKat treated part of the patients’ cohort, CAS provided the post-mortem specimens, FTS treated the majority of the patients and EK participated in the coordination and supervision of the study. All authors approved the final manuscript.