Neuroblastoma is the most common solid, extracranial tumour of childhood and is derived from primitive sympathetic neuroblasts. Amplification of the MYCN oncogene is a strong predictor of poor prognosis and is used worldwide as a parameter for treatment stratification in clinical neuroblastoma trials.1, 2 The MYCN protein is a helix-loop-helix leucine zipper transcription factor expressed during neural crest development. MYCN has been shown to promote cell cycle progression in neuroblastoma cells in vitro3 and transgenic overexpression of MYCN is sufficient to induce neuroblastoma in mice.4 Using high throughput gene expression profiling, distinct sets of MYCN-regulated target genes have been identified in neuroblastoma.5, 6, 7, 8, 9 In addition to numerous genes induced by MYCN, all studies have reported significant numbers of genes to be downregulated. MYCN induces gene transcription via the specific E-box sequences in target gene promoters. Binding of the MYCN:MAX heterodimer to the E-box facilitates gene transcription.10 In contrast, MYCN-induced gene silencing does not require MYCN binding to the E-box. Instead, Myc is recruited to the promoter by the DNA-binding factor, Miz-1. A complex of Myc and Miz-1 bound to the promoter leads to transcriptional silencing, at least in part by recruitment of Dnmt3a.11, 12 The mechanisms by which MYCN silences genes are currently less well understood than those for gene induction.
MicroRNAs (miRNAs) are a class of small, noncoding RNAs (22 nt) that act as negative regulators of protein expression. They promote mRNA degradation and repress mRNA translation by sequence-specific interaction with the 3′-untranslated region of mRNAs.13 Hundreds of human miRNAs have been identified within the last few years and are deposited in the Sanger Centre miRBase database (http://microrna.sanger.ac.uk/sequences/).14, 15 MiRNAs are involved in the regulation of multiple physiological processes including apoptosis, proliferation and differentiation, and there is growing evidence that miRNA dysregulation is involved in the pathogenesis of various diseases, including cancer. High-throughput miRNA expression profiling has identified specific miRNAs that are up- or downregulated in different tumour entities.16 Functional evidence for miRNA involvement in cancerogenesis has also been reported, including inhibition of apoptosis and downregulation of angiogenic inhibitors.17, 18, 19 We hypothesised that miRNA dysregulation may also play a role in the biology of neuroblastoma. Given that MYCN is a major regulator of neuroblastoma tumour biology, we investigated (A) if specific miRNAs are regulated by the MYCN transcription factor in neuroblastoma cells, and (B) if these miRNAs are also differentially expressed between MYCN-non-amplified and MYCN-amplified neuroblastomas in vivo.
Material and methods
Inducible MYCN cell line
cDNAs encoding MYCN fused to an oestrogen responsive domain (ER) of the murine oestrogen receptor and the oestrogen responsive domain alone were cloned into the pWZL vector, which harbours a neomycin cassette for stable expression. The ER domain contains a G525R mutation causing unresponsiveness of the protein to endogenous oestrogen receptor activators. The MYCN-non-amplified neuroblastoma cell line, SK-N-SH-EP (here named as SH-EP), expressing a murine ecotrophic receptor was retrovirally transduced with the pWZLneo-MYCN-ER or with the pWZLneo-ER construct using Phoenix-Eco as a helper cell line. Cells were grown in RPMI-1640 supplemented with 10% heat-inactivated FCS (Sigma, Munich, Germany) and neomycin to select for stable transductants. Single cell clones were isolated via limited dilution. Stable and reproducible expression of the fusion gene was confirmed by western blot analysis.
Western blot analysis
Whole cell lysates, protein electrophoresis and western blotting were conducted according to standard protocols. IMR-32 and SH-EP cell lysates were used as positive and negative controls, respectively.
Cell cycle analysis was performed using propidium iodide (PI, Sigma-Aldrich Inc., Munich, Germany) FACS. Cells were induced with either 200 nM 4-hydroxy-tamoxifen (4-OHT) or ethanol for 72 hr and fixed subsequently. After staining with PI, the DNA content of the cells was analysed.
SH-EP-MYCN-ER cells were induced with 4-OHT for 72 hr. Total RNA was isolated using the TRIzol® reagent (Invitrogen Corp, Carlsbad, CA), and subsequently transcribed into cDNA. PCR primers for human ornithine decarboxylase 1 (ODC1), prothymosin-α (PTMA), nucleolin (NCL) and human telomerase reverse transcriptase (hTERT) were used for quantitative RT-PCR (qRT-PCR, primer sequences are available upon request). Amplicon quantity was determined using SYBR green (F. Hoffmann-La Roche, Grenzach-Wyhlen, Germany) intensity. Gene expression was normalised to β2-microglobulin expression.
Real-time PCR with stem-loop reverse transcription
For validation of the microRNAs (miRNA) microarray experiments, qRT-PCR using a set of 160 Taqman miRNA Assays (Early Access Kit, Applied Biosystems) or Assays on Demand (Applied Biosystems, Foster City, CA miR-17 and miR-92) was performed. Reverse transcription was performed using stem-loop primer optimised for the detection of mature miRNAs as previously described.20
For microarray experiments chips were printed using the mirVana® miRNA Probe Set (Ambion, Austin, TX), which includes probes corresponding to 384 miRNA. Each probe was printed in triplicate. Total RNA was extracted with TRIzol. Tumour samples were derived from stroma-poor tumours and contained at least 80% tumour cells, as estimated by histopathological examination. Tumour RNA was prepared using TRIzol, a FastPrep FP220 Homogenizer and Lysing Matrix D Microbeads (both from Qbiogene, Carlsbad, CA). RNA fragments <40 nt were prepared using the FLASH-PAGE apparatus according to the manufacturer's protocol(Ambion). The mirVana labelling kit (Ambion, http://www.ambion.com/techlib/resources/miRNA/index.html) was used for labelling and hybridisation according to the manufacturer's protocols. A ‘flip colour’ design was used to analyse MYCN-ER induction to exclude dye specific effects. A reference was generated for primary tumours containing pooled RNA from 6 neuroblastoma cell lines (SH-SY5Y, LAN-1, IMR-32, SK-N-AS, SH-EP and IMR-5).
The limma package of the Bioconductor project (http://www.bioconductor.org) was used for data analysis. Median signal intensities and background intensities were obtained for each spot using both channels. To account for spot differences, the background-corrected ratio of the 2 channels was calculated and log2 transformed. The raw data were standardised using global median normalisation to balance fluorescence intensities for the 2 dyes as well as to allow for comparison of expression levels across experiments. As each miRNA was measured in triplicate, the mean log2 ratios (M) of the 3 spots were calculated.
To identify miRNAs differentially expressed in SH-EP MYCN-ER cells treated with 4-OHT as compared with untreated cells, miRNAs having a fold-change difference of at least 2 and at least an absolute value of the t-statistic of 1.96 were selected for further analysis.
Significance Analysis of Microarrays (SAM) was used to select miRNAs differentially expressed between MYCN-amplified and MYCN-non-amplified neuroblastomas. We used the Microsoft Excel Add-In for SAM, which we obtained from http://www-stat.stanford.edu/∼tibs/ SAM (version1.23).21 The one-class data set was analysed using the following parameter settings: 1,000 permutations, k-nearest neighbour imputer with 10 neighbours and 22,471,838 as a seed for the random number generator. A median false discovery rate (FDR) of 0.05 was used as the cut-off for selecting differentially expressed genes and to control for multiple testing.
Establishing an in vitro model of conditional MYCN activation
We first aimed to analyse the effect of MYCN on the miRNA expression profile in neuroblastoma cells in vitro. For that purpose, the MYCN non-amplified neuroblastoma cell line, SH-EP, was stably transfected with an expression vector encoding for MYCN fused to an oestrogen responsive domain mutated to specifically bind only 4-hydroxy-tamoxifen (4-OHT) and not natural oestrogens (Fig. 1a).22, 23 High and stable expression of this fusion protein in 3 independently transduced SH-EP MYCN-ER clones was demonstrated using western blotting (Fig. 1b). Expression levels of the MYCN-ER fusion protein were comparable to MYCN protein expression levels in the MYCN-amplified neuroblastoma cell line, IMR-32 (Fig. 1b). As MYCN:MAX heterodimers are crucial for the proper function of MYCN, we also analysed MAX protein expression. MAX protein levels were comparable in SH-EP MYCN-ER cells and MYCN-amplified IMR-32 cells (Fig. 1c). Activation of MYCN-ER in SH-EP cells induced morphological changes resulting in undifferentiated small, round cells. In contrast, the parental SH-EP cell line and SH-EP cells expressing the ER domain only displayed a more differentiated phenotype with some short dendritic processes both in the presence and absence of 4-OHT (Fig. 2a and data not shown). Cell cycle analysis of nonsynchronised, dividing SH-EP MYCN-ER cells revealed an increased proportion of cells in the S and G2/M phase as well as an increase of apoptotic cells upon 4-OHT-induced MYCN-ER activation (Fig. 1c). This is consistent with previous reports demonstrating MYCN-induced cell cycle progression.3 To functionally validate the SH-EP MYCN-ER model system, we analyzed gene expression of 4 bona fide MYCN target genes 72 hr after 4-OHT treatment: ornithine decarboxylase 1, prothymosin-α, human telomerase reverse transcriptase and nucleolin.3, 24, 25, 26 All 4 genes were induced following 4-OHT treatment of SH-EP MYCN-ER as monitored by qRT-PCR (Fig. 1d). Induction of prothymosin-α was only marginal, while ornithine decarboxylase 1 was induced 3-fold.
Identification of MYCN target miRNAs in vitro
Having established and validated an in vitro model system for ectopic MYCN overexpression, we analysed the global regulation of microRNA expression by MYCN in SH-EP MYCN-ER cells (Fig. 2a). For this purpose, a miRNA microarray was developed using a library containing probes directed against 384 miRNAs.27 The expression of 15 unique miRNAs was induced at least 2-fold (with a t-statistic of at least 1.96) upon addition of 4-OHT to SH-EP MYCN-ER cells. Surprisingly, no miRNA was downregulated (Fig. 2b). None of the 15 MYCN-induced miRNA was regulated upon addition of 4-OHT to the SH-EP-ER control cell line (data not shown). A set of 160 primer pairs was employed to analyse and validate miRNA expression using qRT-PCR with stem-loop reverse transcription in a high-throughput manner. Using this methodology, we validated expression levels of 11 out of the 15 differentially regulated miRNAs. Three miRNAs could not be validated using this method because no real-time assays were available (let-7c, let-7f, miR-93). E-box sequences were detected in the promoter regions of 11 of the 15 regulated miRNAs. The regulation of miR-145 could not be validated using real-time PCR, and no E-box sequence was detected in miR-145. All 3 miRNAs for which no real-time assay existed contained E-boxes, suggesting a good probability that these miRNAs are also regulated by MYCN. As E-boxes are the preferred binding sites of MYCN, these findings suggest direct regulation of miRNAs via MYCN.
To further validate miRNAs induced by MYCN expression in our inducible model system, we analysed expression of miR-17 and miR-92 in MYCN-amplified versus non-amplified cell lines. While expression of miR-17 and miR-92 was low in MYCN-non-amplified SH-EP and SK-N-AS cells, we detected significantly higher levels of miR-17 and miR-92 expression in the MYCN-amplified Kelly, LAN-1 and BE-2-C cell lines (Fig. 2e)
Analysis of primary neuroblastomas
We then used the miRNA microarrays to analyse miRNA expression in 24 pre-treatment primary neuroblastomas of different stages (6 stage 1, 14 stage 4 and 4 stage 4s). None of the stage 1 and 4s tumours, but 7 of the stage 4 tumours had amplified MYCN. Differential expression of 14 miRNAs was recorded between neuroblastomas with or without MYCN amplification according to the SAM algorithm21 at a false discovery rate (FDR) of 0.05. All of these miRNAs were upregulated in the MYCN-amplified tumours. These results are visualised in the heatmap shown in Figure 2c.
Integration of in vitro and in vivo data
Comparison of miRNAs upregulated in both the in vitro culture system overexpressing MYCN and tumours to amplified MYCN revealed that 7 miRNAs were commonly upregulated (Fig. 2d). Quantitative real-time PCR confirmed MYCN mediated regulation of 6 of the 7 miRNAs (Fig. 2b). Interestingly, 4 of these 7 miRNAs, miR-92, miR-106a, miR-17-5p and miR-93, are encoded by 3 paralogous miRNA clusters, the miR-17 cluster on chromosome 13, the miR-106b cluster on chromosome 7 and the miR-106a cluster on the X chromosome.
We here report the results of a comprehensive analysis of miRNA expression in neuroblastoma. We not only used a miRNA microarray for high-throughput analysis of miRNA expression, but also employed a high-throughput implementation of stem-loop reverse transcription real-time PCR to validate our microarray data. The direction of expression change detected using both methods was identical with only one exception, miR-145. The fold-change of miRNA expression detected using either qRT-PCR or the microarray was not identical. Differences in fold-change can most likely be attributed to the different dynamic ranges of both methods.
MYCN amplification is the most prominent genetic alteration of neuroblastoma, being a major determinant of neuroblastoma tumour biology. We found 15 miRNAs that correlated with MYCN amplification in vivo. To link these miRNAs functionally to regulation by MYCN, we complemented the in vivo data with data obtained in our in vitro model of conditionally activated ectopic MYCN, and data obtained from MYCN-amplified and non-amplified cell lines. Seven miRNAs were correlated with MYCN amplification in vivo as well as being regulated upon MYCN activation in vitro. These findings suggest that overexpression of these 7 miRNAs is functionally linked to MYCN amplification in neuroblastoma.
Surprisingly, we identified no miRNAs that were downregulated by MYCN in vitro or inversely correlated with MYCN amplification in vivo. Similar findings to ours for neuroblastoma were reported for c-Myc in a leukemia cell line by O'Donnell. She found that expression of c-Myc resulted in upregulation of several miRNAs but no detectable miRNA repression.28 A recently published report by Chen and Stallings.29 analysed the expression of 157 miRNAs via stem-loop real-time PCR in primary neuroblastomas and in MYCN-amplified neuroblastoma cell lines upon MYCN downregulation by all-trans retinoic acid (ATRA) or siRNA. Four of the 5 miRNAs found by Chen and Stallings to positively correlate with MYCN amplification are among the 15 miRNAs we also identified to positively correlate with MYCN amplification in primary neuroblastomas (let-7b, miR-92, miR-181a, miR-181b). In contrast to our results, these authors also reported a group of miRNAs that were downregulated upon MYCN downregulation and which were inversely correlated with MYCN amplification. The 7 miRNAs we found upregulated in vivo and in vitro by MYCN in neuroblastoma were miR-92, miR-106a, let-7b, miR-17_5p, miR-93, miR-99 and miR-221. Chen and Stallings reported that Let-7b and miR-92 were upregulated upon MYCN downregulation with siRNA in MYCN-amplified neuroblastoma cell lines, although less than 2-fold. They also reported that MYCN downregulation with ATRA treatment resulted in a less than 2-fold let-7b upregulation but a greater than 2-fold miR-92 downregulation, in line with our results for miR-92. These partially conflicting results raise the question, of whether our model system of ectopic MYCN expression can be directly compared to a model system, in which MYCN was downregulated by either ATRA or siRNA. Downregulation of MYCN (by treatment with either ATRA or siRNA) in MYCN-amplified cell lines was shown to induce differentiation and apoptosis,28 whereas MYCN activation (by ectopic overexpression) predominantly induced cell cycle progression and cell growth. Therefore, it is not surprising to detect complementary rather than identical sets of regulated miRNAs. In addition, while Chen and Stallings used only a real-time PCR approach to analyse 157 miRNAs,29 we used a combination of microarray analysis and high-throughput real-time PCR to analyse and validate expression of 384 miRNAs. The reliance on a single analysis platform may be a source of some of the apparently conflicting results between the 2 in vitro model systems used in Chen and Stallings and our results reported here. In our opinion, the holistic biological differences between a model system designed to overexpress NMYC or silence NMYC should lead to the identification of complementary rather than identical sets of genes, and this aspect probably accounts for most discrepancies in these 2 reports.
Consistant with our findings, O'Donnell et al. reported that miR-17-5p, miR-92 and miR-106a were upregulated by c-Myc in B-cells, suggesting that c-Myc regulates the miR-17 and miR-106a clusters.28 When the high level of homology between MYCN and c-Myc is taken into account as well as the fact that MYCN can completely replace c-Myc in a transgenic mouse model,30 it can be expected that MYCN and c-Myc should induce at least similar sets of target miRNAs. We found miR-93, which is encoded in the miR-106b cluster (Figs. 2a–2d), to be upregulated by MYCN. No miRNA of this cluster was found to be regulated by c-Myc.28 This difference may either be caused by the different cellular background or be a distinct difference between MYCN and c-Myc miRNA regulation.
Several lines of evidence point to the oncogenic properties of the miRNAs of the miR-17 and miR-106a clusters. Overexpression of miR-17-5p, miR-92 and miR-106a is a general feature of carcinomas,16, 31 and the miR-17 cluster is both amplified and overexpressed in lymphomas.32, 33 Overexpression of the miR-17 cluster in haematopoietic stem cells led to lymphoma formation in mice,34 providing functional evidence for the involvement of those miRNAs in lymphomagenesis. In addition, the miR-17 miRNA cluster induced by c-Myc was demonstrated to enhance angiogenesis by downregulating angiogenic inhibitors in a murine colon tumour model.35
Additionally, we found and validated miR-221 to be induced by MYCN in vitro, and to be differentially regulated between MYCN amplified and non-amplified neuroblastomas. This miRNA has not been linked to the Myc family of transcription factors before, suggesting that it is either a miRNA targeted specifically by MYCN or by MYCN in the context of neuroblastoma. Overexpression of miR-221 has been detected in papillary thyroid carcinoma,36 glioblastoma37 and pancreatic cancer,38 and might well play a role in the tumorigenesis of several tumour entities.16
In addition to the potential role of MYCN-regulated miRNAs in neuroblastoma pathogenesis, miRNA induction by MYCN might be a general mechanism of MYCN-mediated protein regulation. Previous studies have shown that c-Myc downregulates gene expression via transcriptional repression by the MYC:Miz-1 complex.12 Recently, the Myc:Miz-1 complex was reported to also incorporate Dnmt3a, thereby, facilitating DNA methylation and constitutive transcriptional silencing.11 Translational silencing mediated by MYCN induced miRNA upregulation is an independent additional mechanism by which MYCN could downregulate gene expression.
Deregulation of miRNAs may be one feature contributing to the heterogeneity of neuroblastoma. The functional implications of miRNA deregulation in neuroblastoma, as well as evaluation of their potential application as novel targets for therapy using ‘antagomirs’39 deserves attention in future studies. Here we provide evidence of differential miRNA expression in primary neuroblastomas, and identify a MYCN-regulated subset of these miRNAs. Additionally, we present that miRNA induction may be a new mechanism for MYCN to downregulate gene expression.
We thank Ambion, namely Dr. M. Canneux, for excellent product support and Dr. M. Serrano for providing the pWZL-neo vector. This work was supported by grants from the German National Genome Research Network (BMBF/NGFN2) to A.E., A.S. and M.E., by grants from the Sanitätsrat Dr. Hüber und Gemahlin Stiftung, the Bohne Junius Stiftung and the Vigoni/DAAD program to AE, and by a grant from the Stiftung P.E. Kempkes to B.B. and H.C.