Correspondence to: Johannes H. Schulte, MD, Pediatric Oncology and Hematology, University Children's Hospital Essen, Hufelandstr. 55, 45147 Essen, Germany, Tel.: +49–20172385185], Fax: +49–201-723–5750, E-mail: Johannes.firstname.lastname@example.org
Neuroblastoma is the most common extracranial solid tumor of childhood, and accounts for ∼15% of all childhood cancer deaths. The histone demethylase, lysine-specific demethylase 1 (KDM1A, previously known as LSD1), is strongly expressed in neuroblastomas, and overexpression correlates with poor patient prognosis. Inducing differentiation in neuroblastoma cells has previously been shown to down regulate KDM1A, and siRNA-mediated KDM1A knockdown inhibited neuroblastoma cell viability. The microRNA, miR-137, has been reported to be downregulated in several human cancers, and KDM1A mRNA was reported as a putative target of miR-137 in colon cancer. We hypothesized that miR-137 might have a tumor-suppressive role in neuroblastoma mediated via downregulation of KDM1A. Indeed, low levels of miR-137 expression in primary neuroblastomas correlated with poor patient prognosis. Re-expressing miR-137 in neuroblastoma cell lines increased apoptosis and decreased cell viability and proliferation. KDM1A mRNA was repressed by miR-137 in neuroblastoma cells, and was validated as a direct target of miR-137 using reporter assays in SHEP and HEK293 cells. Furthermore, siRNA-mediated KDM1A knockdown phenocopied the miR-137 re-expression phenotype in neuroblastoma cells. We conclude that miR-137 directly targets KDM1A mRNA in neuroblastoma cells, and activates cell properties consistent with tumor suppression. Therapeutic strategies to re-express miR-137 in neuroblastomas could be useful to reduce tumor aggressiveness.
Neuroblastoma is the most common extracranial childhood tumor, accounting for >15% of all childhood cancer deaths. The clinical hallmark of neuroblastoma is its heterogeneity. Neuroblastomas with favorable biology (Stages 1, 2 and 4s) often undergo complete regression or differentiation even without therapy, while high-stage, highly aggressive neuroblastoma often ends fatally despite recent therapeutic improvements. Several biological markers have been identified defining high-risk neuroblastoma, such as amplification of the MYCN oncogene, low expression of the NTRK1 neurotrophic tyrosine kinase receptor and LOH of chromosome 1p or 11q.[2, 3] MYCN amplification occurs in >30% of high-stage neuroblastomas and only in 5% of Stage 1, 2 and 4s tumors. In contrast, 30% of neuroblastomas express high NTRK1 levels, most prominently occurring in Stage 1 tumors, and this correlates with good patient prognosis.[4, 5]
Lysine (K)-specific demethylase 1A (KDM1A, previously known as LSD1) is a histone demethylase, which specifically catalyzes the demethylation of di- and monomethylated lysine 4 in histone 3 through amine oxidation. KDM1A is highly conserved in eukaryotes, and targeted disruption of the Kdm1a gene locus in mice results in embryonic lethality. Overexpression of KDM1A has been shown in a variety of human cancers, including prostate, bladder, breast, colorectal, gastric and lung cancers, neuroendocrine carcinomas several sarcomas and most recently acute myeloid leukemia (AML).[8-14] KDM1A is strongly expressed in undifferentiated neuroblastomas, and overexpression correlates with poor patient prognosis. Inducing differentiation in neuroblastoma cells in vitro downregulates KDM1A expression, and siRNA-mediated KDM1A knockdown or inhibition with small molecular inhibitors induces growth inhibition of NB cells.
MicroRNAs (miRNAs) are small noncoding double-stranded RNA molecules, 21–22 nucleotides in length that regulate 30% of human gene expression at a posttranscriptional level by targeting messenger RNAs (mRNAs).[16, 17] MiRNAs bind to mRNA in the 3′-untranslated region (UTR) resulting in translational repression or cleavage and degradation of mRNA. Furthermore, each miRNA can regulate multiple mRNAs and each mRNA can be targeted by a number of miRNAs. In cancer, the function of miRNAs is dependent on their mRNA targets, so they can act as tumor suppressors or as oncogenes. In primary neuroblastomas, miRNA expression profiles correlate with 11q deletion, hyperdiploidy, MYCN amplification and prognosis.[19-22] The MYCN-regulated polycistronic miR-17–92 cluster (miR-17-5p, −18a, −19a, −20a and −92) acts as an oncogene in many types of cancer, including neuroblastoma.[18, 23] The oncogenic function of miR-17–92 was confirmed in neuroblastoma cell lines by identifying targets, mainly within TGF-beta signaling. Thus, miRNAs are thought to play essential roles in cell-cycle control, cell adhesion and cell death in cancer. MiR-137 is located on chromosome 1p22 within the gene for the AK094607 non-protein-coding RNA, and is downregulated in several human cancer types, including colorectal cancer, glioblastoma, oral cancer and squamous cell carcinoma of the head and neck.[26, 27] In this study, we analyzed the role of miR-137 in neuroblastoma biology and its interplay with KDM1A, a predicted target of miR-137.
Material and Methods
The human neuroblastoma cell lines, IMR-32, SHEP and SK-N-BE and the HEK-293 human embryonic kidney cell line were cultivated in RPMI1640 medium supplemented with 10% FCS, amphotericin B (3 µg/ml), penicillin (100 U/ml) and streptomycin (100 µg/ml) and maintained at 37°C in a humidified 5% CO2 incubator. The identity of all cell lines was verified by the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany)
Transient agomir and siRNA transfection
The SiPort™ NeoFX™ transfection agent (Applied Biosystems, Darmstadt, Germany) was used to transiently transfect 30 nM of the Cy™3-labeled Pre-miR™ negative control (Applied Biosystems) or the Pre-miR™ miRNA-137 precursor molecules (Applied Biosystems) into neuroblastoma cell lines or HEK-293 cells according to the manufacturer's instructions. FlexiTube siRNA-HsLSD1 (Qiagen, Hilden, Germany, #SI02780932), AllStars negative control siRNA (Qiagen, #1027281), Silencer® Select Validated siRNAs (Applied Biosystems, #4390824) for KDM1A_1 (#s617) or KDM1A_2 (#s618) or Silencer® Negative Control No. 1 siRNA (Applied Biosystems, #AM4611) was transiently introduced in a final concentration of 33 nM into neuroblastoma cell lines by reverse transfection using the Lipofectamine™ 2000 transfection reagent (Invitrogen, Darmstadt, Germany) following the manufacturer's protocol.
Inducible expression of KDM1A lacking the 3′UTR
For rescue experiments, SHEP cells were electroporated with pcDNA™6/TR (Invitrogen) and either pT-Rex™-DEST30-GFP or pT-Rex™-DEST30-KDM1A (Invitrogen), encoding GFP or KDM1A, respectively. Transfected cells were selected by blasticidin (7.5 mg/ml) and geneticin (0.5 mg/ml), and single cell clones were obtained by limited dilution. GFP or KDM1A expression was induced in transfected cells by adding 1 µg/ml tetracycline to the selection medium.
Apoptosis, cell proliferation and cell viability assays
Cells were seeded onto 96-well plates and transfected with Pre-miRs or siRNA. Apoptosis was measured using the Cell Death Detection ELISAPLUS (Roche, Mannheim, Germany), and cell proliferation was assessed by BrdU incorporation using the Cell Proliferation ELISA, BrdU (Roche) following the manufacturer's protocols. Metabolic activity was analyzed using the 3-[4,5-dimethylthiazol-3-yl]-2,5-diphenyltetrazolium bromide (MTT) assay (Roche). Based on the assumption that the metabolic activity is constant in a given experiment, we use the metabolic activity measured in MTT assays as a surrogate for the number of live cells, thus, cell viability. All experiments were performed in triplicate and repeated independently at least three times.
Cell cycle analysis
To monitor changes in the cell cycle distribution, cells were harvested 96 hr after transfection, washed with PBS, fixed in 70% ethanol and stained with 10 µg/ml propidium iodide. The cellular DNA content was analyzed on a FC500 Flow Cytometer (Beckman Coulter, Krefeld, Germany). Experiments were performed in triplicate.
Luciferase reporter assays
HEK-293 and SHEP cells were cotransfected with 1 µg/ml of the pMirTarget vector containing the target site of KDM1A 3′-UTR (OriGene, Rockville), 0.2 µg/ml pRL-TK vector (Promega, Mannheim, Germany) as an internal control and 20 nM of the Pre-miR using DharmaFECT duo (Thermo Scientific, Schwerte, Germany). Firefly and renilla luciferase activities were measured 48 h after transfection using the Dual-Luciferase® Reporter Assay System (Promega) according to the manufacturer's instructions.
Gene expression analysis
Total mRNA was isolated from cells using the RNeasy® Micro Kit (Qiagen), and cDNA reverse-transcribed by Super Script II Reverse Transcriptase (Invitrogen). Real-time PCR was performed using TaqMan® Fast Advanced Master Mix (Applied Biosystems) and the StepOnePlus™ Real-Time PCR System (Applied Biosystems) according to the manufacturer's instructions. KDM1A, NEFL and TFPI2 mRNA expression was normalized to human endogenous GAPDH expression and calculated using the ΔΔ-Ct method. All experiments were performed in triplicate and repeated independently at least three times.
Western blot analysis
Cells were lysed 30 min on ice in RIPA buffer (50 mM HEPES, 10 mM NaCl, 0.1% SDS, 1% Triton X-100, 1% NP-40, Roche Complete Protease Inhibitor Cocktail and Roche PhosSTOP Phosphatase Inhibitor Cocktail). Proteins (50 µg) were separated on 10% SDS-PAGE and transferred to Amersham Hybond™-C Extra (GE Healthcare, Solingen, Germany) membranes. Nitrocellulose membranes were blocked in 5% milk powder in TBS-T0,1 and incubated with the primary antibodies against KDM1A (1:1,000, #2139S, Cell Signaling, Frankfurt am Main, Germany) or caspase 9 (Immunotech, #PN IM3434) and GAPDH (1:2,000, #MAB374, Millipore, Darmstadt, Germany) as a loading control, followed by anti-rabbit-HRP (1:2,000, #NA9340, GE Heathcare) and anti-mouse-HRP (1:5,000, #ab97046, Abcam, Berlin, Germany) for 1 hr at room temperature. Proteins were visualized using Amersham ECL Plus™ Western Blotting Detection Reagents (GE Healthcare) and analyzed on UVchem Detection Device (Biometra, Göttingen, Germany).
Statistical analyses were conducted using SPSS Version 19 or Graph Pad Prism 5.0. Kaplan-Meier analyses were used to analyze event-free (EFS) and overall survival (OS). Expression of miRNA was normalized as previously described.[19, 20] NTRK1 and miR-137 expression were used as continuous variables, and high or low expression levels were defined as expression above or below the median. A two-sided Wilcoxon rank test was used to test for differential miR-137 expression. KDM1A, NEFL and TFPI2 expression levels were calculated using biogazelle (www.bigazelle.com). The significance of differential metabolic activity (as a surrogate for cell viability) over time was assessed by a permutation test, which simulated the null hypothesis that the observed differential metabolic activity was generated by random noise. Logarithmic fold-change between the mean values of both treatments and the negative control were calculated for each timepoint. The absolute sum of these fold-changes was used as the differential metabolic activity score. Observed values from all timepoints and replicates were randomly permuted 100,000 times, and p-values were calculated as the fraction of all permutations with a score at least as high as the score for the original observed values. The significance of differences between other experimental treatment groups was assessed by two-sided student's t-test and indicated by *(p < 0.05), **(p < 0.01) and ***(p < 0.001).
miR-137 is differentially expressed in neuroblastomas
We assessed miR-137 expression in 69 primary neuroblastomas previously profiled for miRNA expression. This cohort was representative in terms of clinical and molecular covariates (supporting information Table 1) and their correlation with EFS and OS (supporting information Fig. 2). Kaplan-Meier analysis revealed that patients with tumors expressing high levels of miR-137 had better predicted EFS (p = 0.029) and OS (p = 0.033) than patients with tumors expressing low miR-137 levels (Fig. 1a). Consistently, Stage 3 and 4 tumors expressed significantly higher levels of miR-137 than tumors with the Stages 1, 2 (p = 0.00005) or 4S (p = 0.009). MYCN-amplified tumors also showed a tendency to express lower miR-137 levels than tumors with single-copy MYCN. This tendency did not reach significance in the primary cohort, possibly due to the low number of NMYC-amplified tumors. However, miR-137 expression was significantly lower in MYCN-amplified tumors as compared with tumors with single-copy MYCN in both validations cohorts (p = 0.01 and p = 3e-11). Significantly higher miR-137 levels were detected in neuroblastomas with high NTRK1 expression levels (Fig. 1b, p = 0.009). These findings were validated by re-analysis of previously published expression data from independent cohorts of 132 and 105 primary neuroblastomas (supporting information Table 1, supporting information Fig. 1 and 2). To further compare absolute miR-137 expression levels with those of other miRNAs, we reanalyzed miRNA transcriptome sequencing data from the same tumors (supporting information Table 2). As expected, miR-137 expression was lower in unfavorable than favorable neuroblastomas (p = 0.046). However, even in favorable neuroblastomas, absolute miR-137 expression levels were much lower than the absolute expression of the oncogenic miRNAs, miR-17 (p = 4e-10) and miR-92a (p = 2e-8) (Fig. 1c). Taken together, we demonstrate a correlation between high miR-137 expression and favorable tumor biology in neuroblastomas, suggesting a tumor-suppressive function for this miRNA.
miR-137 represses neuroblastoma cell viability
To assess miR-137 function in connection with tumor cell characteristics, we analyzed metabolic activity by MTT assay as a surrogate measure for cell viability after re-expressing miR-137 in three human neuroblastoma cell lines. Re-expression of miR-137 significantly reduced total cellular metabolic activity over time in all cell lines analyzed. Cultures observed microscopically did not grow as fast and this reduction in cell viability was confirmed in MTT assays (Figs. 2a and 2b). Cell cycle analysis (supporting information Fig. 3a) revealed an increase in the subG1 cell fraction of all three cell lines by miR-137 re-expression (Fig. 2c), indicating an increase in apoptotic cell death. This effect was most pronounced in SHEP cells, but was significant for all three cell lines analyzed (supporting information Fig. 3c). We confirmed this miR-137-induced apoptosis using an ELISA to quantify cell death in cultures compared with cells transfected with scrambled miRNA controls (Fig. 2e). This proapoptotic effect of miR-137 was further substantiated by the fact that caspase 9 cleavage was induced by miR-137 expression (Fig. 2f). Increased apoptosis was attended by diminished cell proliferation, as indicated by lower BrdU incorporation levels for miR-137-expressing cells (Fig. 2d). To test whether miR-137 facilitates neuronal differentiation, we analyzed neurofilament (NEFL) expression as a marker of neuronal differentiation. Indeed, miR-137 expression resulted in a significant induction of NEFL expression in all analyzed cell lines, indicating that miR-137 induces neuronal differentiation in neuroblastoma cells (Fig. 2g). These experiments demonstrate that miR-137 reduces neuroblastoma cell viability and proliferation and induces apoptosis and neuronal differentiation, in line with a tumor-suppressive function for this miRNA in neuroblastoma.
miR-137 regulates KDM1A expression in neuroblastoma cells
In search of potential target genes of miR-137, we made use of the TargetScan software, which predicted KDM1A as a miR-137 target (Fig. 3a). We aimed to validate this predicted interaction functionally using reporter assays as well as by re-expression of miR-137 in neuroblastoma cell lines. Expression of KDM1A at both the mRNA and protein levels was reduced upon miR-137 re-expression in the SHEP, IMR-32 and SK-N-BE neuroblastoma cell lines (Figs. 3b and 3c). To confirm a direct binding of miR-137 to target regions within the KDM1A gene a reporter assay was conducted in both the SHEP neuroblastoma cell line as well as the HEK-293 cell line, a cell line frequently used for reporter assays. A luciferase reporter plasmid carrying the 3′-UTR of KDM1A, the plasmid pRL-TK, which encodes renilla luciferase as an internal control and the miR-137 expression vector were co-transfected into HEK-293 and SHEP cells. This resulted in decreased luciferase activity (normalized to renilla activity) compared with controls (Fig. 3d), confirming the direct interaction between miR-137 and the 3′-UTR of the coding gene for KDM1A in neuroblastoma and HEK-293 cells. We have previously shown that the expression of the gene encoding tissue factor pathway inhibitor 2, TFPI2, was induced upon KDM1A knockdown by siRNA in neuroblastoma cells. As expected, qPCR of TFPI2 expression in neuroblastoma cells re-expressing miR-137, and thus, with downregulated KDM1A, also showed induction of TFPI2 (Fig. 3e). Taken together, these experiments demonstrate that KDM1A is directly downregulated by miR-137, and that KDM1A downregulation by miR-137 induced expression of the KDM1A-regulated gene, TFPI2.
KDM1A knockdown phenocopies the effect of miR-137 re-expression
We assessed the degree to which KDM1A downregulation contributed to the effects of miR-137 re-expression on cell viability, proliferation and induction of apoptosis using siRNA-mediated knockdown of KDM1A. Targeting KDM1A with siRNA downregulated KDM1A protein expression in all three neuroblastoma cell lines (Fig. 4a). We then determined neuroblastoma cell viability after transfection of siRNA against KDM1A mRNA. MTT assay revealed a significant reduction in total metabolic activity (surrogate measure for cell viability) over time upon siRNA-mediated KDM1A knockdown (Fig. 4b). Flow cytometric analysis (supporting information Fig. 3b) also revealed a high proportion of apoptotic cells (Fig. 4c, supporting information Fig. 3d, supporting information Table 3). Notably, IMR-32 cells also responded to KDM1A/LSD1 knockdown with a strong block in proliferation (Fig. 4d), while cell death was the more pronounced reaction to miR-137 re-expression (Fig. 2c). Consistently, siRNA against KDM1A reduced cell proliferation as determined by BrdU proliferation assay, and increased apoptosis as demonstrated by cell death ELISA (Figs. 4d and 4e). These effects of KDM1A knockdown by siRNA (siKDM1A) were replicated using two different siRNAs directed against KDM1A (siKDM1A_1 and siKDM1A_2; supporting information Fig. 4), indicating that the phenotype is specific to a KDM1A knockdown and not an unspecific effect of the scrambled siRNA used. Taken together, KDM1A knockdown reproduced the effects of miR-137 re-expression and repressed neuroblastoma cell viability and proliferation while inducing apoptosis.
The presence of miR-137-resistant KDM1A partially rescues the effect of miR-137 expression
To determine whether KDM1A is the predominant target of miR-137 and responsible for the effect of miR-137 expression in neuroblastoma cells, we generated the SHEP-KDM1A#1 cell line. Under tetracycline treatment, SHEP-KDM1A#1 cells inducibly express a cDNA for KDM1A lacking the 3′-UTR, which harbors the miR-137 binding site. This KDM1A cDNA is, therefore, resistant to repression by miR-137. Expression of miR-137 in SHEP-KDM1A#1 cells in the absence of tetracycline repressed KDM1A (Fig. 5a) and significantly reduced the total metabolic activity (surrogate measure for cell viability) determined by MTT assay (Fig. 5b). The higher level of KDM1A expression was maintained upon miR-137 expression when cells were treated with tetracycline, thus, expressing the miR-137-resistant KDM1A (Fig. 5a). This was accompanied by a small but significant decrease in total cellular metabolic activity in MTT assays, compared with miR-137 expression in uninduced SHEP-KDM1A#1 cells (71% +/−5% vs. 58% +/−4%, p = 0.04, Fig. 5b). Expression of miR-137 in the SHEP-GFP control cell line, which was generated for tetracycline-inducible expression of GFP using the same vector system, repressed KDM1A expression and cell viability (total cellular metabolic activity in MTT assay) to similar degrees in the presence and absence of tetracycline (Figs. 5a and 5b). These data demonstrate that a certain proportion of the effect of miR-137 expression in neuroblastoma cells occurs via KDM1A downregulation, but that the major proportion is not a singular effect of KDM1A suppression.
MicroRNAs play roles in many aspects of tumor biology. Here, we report that low tumor expression levels of miR-137 are correlated with poor survival of neuroblastoma patients, suggesting a tumor-suppressive function for this miRNA. We demonstrate that miR-137 induces apoptosis and decreases viability and proliferation of neuroblastoma cells in vitro, and provide evidence that these effects are mediated, at least in part, by direct downregulation of KDM1A by miR-137. Since each miRNA regulates multiple mRNAs, and several other target genes have been predicted for miR-137, it is important how much of the complete effect of miR-137 re-expression in neuroblastoma cells occurs via KDM1A or through modulation of other miR-137 targets. We show that KDM1A knockdown phenocopies the effect of miR-137 re-expression on the endpoints assessed here, which are central to tumor cell biology, namely, proliferation, viability and induction of apoptosis. These results support KDM1A as a central miR-137 target gene important in regulating neuroblastoma tumor biology. To determine to which extent other miR-137 target genes contribute to the observed phenotype, we performed a rescue experiment by inducibly overexpressing a miR-137-resistant KDM1A cDNA. The phenotype was only partially rescued by the miR-137-resistant KDM1A cDNA. Taken together with the finding that KDM1A knockdown resulted in a near complete phenocopy of the effect of miR-137, we postulate that KDM1A is a central for the oncogenic actions of miR-137 in neuroblastoma cells, but not the only relevant miR-137 target. This model can also explain the presence of some phenotypic differences between miR-137 re-expression and siRNA-mediated KDM1A knockdown, such as the different proportions of IMR-32 cells to initiate apoptosis or a block in proliferation.
The KDM1A protein acts as an oncogene in neuroblastomas and several other tumors by epigenetically reprogramming gene expression patterns. However, as KDM1A does not bind DNA directly, but is directed to DNA by interaction with different transcription factors, the functional consequences of aberrant KDM1A expression is cell type-specific. Interestingly, KDM1A is also expressed in neural precursor cells and interacts with the NR2E1 transcription factor in these cells. NR2E1, previously known as TLX, is a unique nuclear receptor expressed exclusively in the embryonic brain neuroepithelium, where it represses miR-137 by binding to 5′ and 3′ cis-regulatory sequences. As miR-137 downregulates KDM1A, which is a TLX transcriptional co-repressor, increased miR-137 expression in neural stem cells consequently led to reduced cell proliferation and accelerated neural differentiation. This is in line with the tumor-suppressive effects we observed in neuroblastoma cells following miR-137 re-expression. Mechanisms deregulating KDM1A expression during tumor development have not yet been described in neuroblastoma. KDM1A has recently been described as a miR-137 target in colorectal cancer, where epigenetic silencing of miR-137 in colorectal cancer cells led to aberrant KDM1A expression. While we did not address the mechanism of miR-137 repression in aggressive neuroblastomas, we demonstrated that miR-137 re-expression downregulates KDM1A in neuroblastoma cells, resulting in strong tumor-suppressive effects.
Singular miRNAs are known to target several mRNAs, and their effect on cells is often caused by targeting multiple oncogenes or tumor suppressor genes. Accordingly, miRNA-based therapies may be less likely to cause the development of therapy resistance than conventional and targeted drugs. The treatment of neuroblastoma patients with miR-137 mimics could, therefore, be an interesting approach to reduce tumor aggressiveness. In addition, successful delivery of miRNAs using GD2-coated or plain nanoparticles has been demonstrated to be effective against neuroblastomas in mouse models.[31-33] DNA methylation has been reported to silence miR-137 in colon and oral cancers.[26, 29] The mechanism downregulating miR-137 in the majority of neuroblastomas remains to be elucidated, however, neuroblastoma cells may similarly use methylation. If so, treatment of neuroblastoma cells with DNA demethylating agents, such as 5-azacytidine, could be an alternative approach to re-express miR-137.
Taken together, our results establish a functional link between miR-137 and KDM1A expression in neuroblastoma, demonstrating that KDM1A is directly repressed by miR-137, which is expressed at lower levels in aggressive neuroblastomas. Restoring miR-137 function could represent an alternative approach to reduce therapeutically KDM1A expression, thereby attenuating aggressive tumor properties in vivo.
The authors thank S. Dreesmann for excellent technical assistance and the WTZ Research Support Service (supported in part by the Deutsche Krebshilfe Comprehensive Cancer Center financing) for proofreading the manuscript.