Impact of DLK1-DIO3 imprinted cluster hypomethylation in smoker patients with lung cancer

Oncotarget4395 www.impactjournals.com/oncotarget www.impactjournals.com/oncotarget/ Oncotarget, 2018, Vol. 9, (No. 4), pp: 4395-4410 Impact of DLK1-DIO3 imprinted cluster hypomethylation in smoker patients with lung cancer Sonia Molina-Pinelo1,2,3, Ana Salinas1, Nicolás Moreno-Mata4, Irene Ferrer1,2,3, Rocío Suarez1,2,3, Eduardo Andrés-León1, Manuel Rodríguez-Paredes5,6, Julian Gutekunst5, Eloisa Jantus-Lewintre7,8, Carlos Camps9,10, Amancio Carnero1,3 and Luis Paz-Ares1,2,3 1Instituto de Biomedicina de Sevilla (IBIS) (HUVR, CSIC, Universidad de Sevilla), Sevilla, Spain 2Medical Oncology Department, Hospital Universitario Doce de Octubre & Centro Nacional de Investigaciones Oncológicas (CNIO), Madrid, Spain 3CIBER de Cáncer, Madrid, Spain 4Thoracic Surgery Department, Hospital Universitario Virgen del Rocio, Sevilla, Spain 5Division of Epigenetics, DKFZ-ZMBH Alliance, German Cancer Research Center, Heidelberg, Germany 6University Tumor Center Düsseldorf, University of Düsseldorf, Düsseldorf, Germany 7Molecular Oncology Laboratory, Fundación para la Investigación del Hospital General Universitario de Valencia, Valencia, Spain 8Department of Biotechnology, Universitat Politècnica de Valencia, Valencia, Spain 9Department of Medicine, University of Valencia, Valencia, Spain 10Department of Medical Oncology, Hospital General Universitario de Valencia, Valencia, Spain Correspondence to: Sonia Molina-Pinelo, email: [email protected] Luis Paz-Ares, email: [email protected] Keywords: lung cancer, epigenetic, DLK1-DIO3 cluster, transcriptional regulation, COPD Received: September 16, 2015 Accepted: June 02, 2016 Published: July 15, 2016 Copyright: Molina-Pinelo et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License 3.0 (CC BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. ABSTRACT DNA methylation is important for gene expression and genome stability, and its disruption is thought to play a key role in the initiation and progression of cancer and other diseases. The DLK1-DIO3 cluster has been shown to be imprinted in humans, and some of its components are relevant to diverse pathological processes. The purpose of this study was to assess the methylation patterns of the DLK1-DIO3 cluster in patients with lung cancer to study its relevance in the pathogenesis of this disease. We found a characteristic methylation pattern of this cluster in smoking associated lung cancer, as compared to normal lung tissue. This methylation profile is not patent however in lung cancer of never smokers nor in lung tissue of COPD patients. We found 3 deregulated protein-coding genes at this locus: one was hypermethylated (DIO3) and two were hypomethylated (DLK1 and RTL1). Statistically significant differences were also detected in two different families of SNORDs, two miRNA clusters and four lncRNAs (MEG3, MEG8, MEG9 and LINC00524). These findings were validated using data from the cancer genome atlas (TCGA) database. We have then showed an inverse correlation between DNA methylation and expression levels in 5 randomly selected genes. Several targets of miRNAs included in the DLK1-DIO3 cluster have been experimentally verified as tumor suppressors. All of these results suggest that the dysmethylation of the imprinted DLK1-DIO3 cluster could have a relevant role in the pathogenesis of lung cancer in current and former smokers and may be used for diagnostic and/or therapeutic purposes. Research Paper Oncotarget4396 www.impactjournals.com/oncotarget INTRODUCTION Lung cancer is a highly incident disease, commonly associated to tobacco exposure, and currently represents the most frequent cause of cancer mortality. More than 80% of the cases present non-small cell lung cancer (NSCLC) histology, including adenocarcinoma and squamous cell carcinoma [1, 2]. Many molecular analyses have shown the complexity driving the carcinogenesis of the clinically relevant lung cancer phenotypes [3]. This includes DNA methylation patterns that undergo complex changes in cancer [4]. This process is important for gene expression and genome stability, and its disruption is thought to play a key role in the initiation and progression of lung cancer [5]. According to classical Mendelian laws, the majority of genes in a human cell are inherited in two functionally equivalent parental copies. However, there is a small subset of genes with one turned-off copy in a parent-of- origin-dependent manner[6]. This phenomenon, known as genomic imprinting, is an epigenetic process that involves monoallelic expression. Some genes are imprinted across the complete genome sequence, but the vast majority of imprinted genes are clustered [7, 8]. Genomic imprinting occurs in many diseases, such as neurological disorders, cancer and many others [9-11]. Loss of imprinting is commonly observed in human tumors [12], suggesting the presence of the imbalanced expression of imprinted tumorsuppressor genes or oncogenes. The IGF2 and P73 genes have been pointed out as examples of imprinted oncogene and tumor suppressor gene, respectively, in several tumors [11, 13-20]. The DLK1-DIO3 imprinted cluster is located on chromosome 14q32.2. This region includes protein-coding genes (DLK1, RTL1 and DIO3), long non-coding RNAs (MEG3, MEG8, MEG9 and LINC00524), two large clusters of miRNAs, two families of small nucleolar RNAs (SNORD113 and SNORD114) and several pseudogenes [21]. Aberrations involving some components of the DLK1-DIO3 cluster have been linked to pathological processes [22]. Increased DLK1 levels may become oncogenic in NSCLC by upregulating cell cycle machinery, which has been proposed to be highly associated with tumor invasion [23]. Similarly, overexpression of the RTL1 gene promotes hepatocarcinogenesis [24]. miRNAs located in the DLK1-DIO3 cluster have been shown to be involved in the development of different tumors such as lymphomas, glioblastomas, gastric tumors, and many others [22]. At the same time, it has been demonstrated that SNORD113 expression is linked to decreased survival in patients with hepatocellular carcinoma [25]. Concerning lung cancer, Valdmanis et al. recently described that the expression of a cluster of ~53 microRNAs and mRNAs at the DLK1- DIO3 locus on mouse chromosome 12qF1 was markedly increased in tumors compared to non-tumoral tissue, and they have proposed a role for the cluster in the origin of cancer cells [26]. Altogether, the aforementioned data suggest aberrant epigenetic regulation in genomic imprinting may promote imbalanced growth, thus leading to human tumorigenesis. To elucidate the role of the DLK1-DIO3 cluster in the tumorigenesis of nonsmall cell lung cancer, we have analyzed the methylation pattern in lung tumors as compared to non-tumoral lung tissue. RESULTS DNA methylation pattern of the DLK1-DIO3 cluster in lung cancer To evaluate the potential role of the DLK1-DIO3 cluster in lung cancer, we analyzed the methylation status of the cluster in human lung tissue. The methylation profile of DLK1-DIO3 was evaluated in human tumor samples and compared to non-tumoral tissue using the Illumina Infinium Human Methylation 450 BeadChip. The methylation levels in lung cancer versus non-tumoral tissue are represented in Figure 1A and Supplementary Table S1. Patients with lung cancer showed three deregulated protein-coding genes: one was hypermethylated (DIO3) and two were hypomethylated (DLK1 and RTL1). Statistically significant differences (adjusted p-value < 0.05) were also detected in two different families of SNORDs (SNORD113 and SNORD114), with the exception of SNORD114-30 that showed a similar but non-significant pattern of DNA hypomethylation (p=0.07), (Table 1). Similarly, two miRNA clusters and four lncRNAs (MEG3, MEG8, MEG9 and LINC00524) were found to be significantly less methylated in tumors as compared to non-tumoral tissue (Figure 1B and Table 1). To validate these findings we used methylation data from an independent cohort obtained from TCGA database, and a different method of analysis. Consistently significant differences in methylation were observed again for the DLK1-DIO3 cluster (Figure 2). All genes included in the cluster show a hypomethylated pattern with the exception of DIO3 that was hypermethylated in tumor tissue. In order to investigate methylation-dependent RNA expression changes, some components of the DLK1- DIO3 cluster, such as DIO3, SNORD113-5, SNORD113-7, SNORD114-9, and miR-889, were randomly selected to analyze their expression levels in matched normal-tumor samples. The results confirmed that the methylation levels were inversely associated with the expression levels, i.e., hypomethylation of SNORD113-5, SNORD113-7, SNORD114-9, and miR-889 correlated with high gene expression in tumor samples relative to non-tumoral tissue. In the case of DIO3, we found lower levels of mRNA expression in lung tumor tissue (Figure 3). Oncotarget4397 www.impactjournals.com/oncotarget DLK1-DIO3 methylation profile by smoking characteristics We evaluated variations of the DLK1-DIO3 methylation profile of lung tissues, tumoral and nontumoral, according to the prior smoking exposure of patients. We found no differences on cluster DLK1-DIO3 methylation between non-tumoral and tumoral lung tissue among never smokers (Figure 4A). The methylation status of DIO3, miR-889 and SNORD114-9 was further validated by 454 bisulfite sequencing using pooled non-tumor and tumor samples from non-smokers (Figure 4B). In current and former smokers with lung cancer, the DLK1-DIO3 cluster showed higher methylation in non-tumoral tissue as compared to tumor. In the latter group of patients, statistically significant differences in methylation were identified in all components of the DLK1-DIO3 cluster (adjusted p-value< 0.001) (Figure 4C). We then next analyzed whether the DNA methylation pattern in non-tumoral lung parenchyma of current and former smokers with COPD differed from that of current and former smokers without COPD and found no statistical differences (Figure 1B). Of note, an increase in the DNA promoter methylation of DIO3 was noticeable in COPD lung tissue although not statistically significant (log 2 ratio COPD + /COPD - LC - was 0.535 and the adjusted p-value was 0.840) (Supplementary Table S1). In addition, we have found that the DLK1-DIO3 methylation profile of non-malignant tissue from lung cancer patients and control subjects was similar (data not shown). Figure 1: Methylation profile of the DLK1-DIO3 cluster in lung cancer. A. The DLK1-DIO3 cluster is located on chromosome 14 (14q32). Using a Circos plot, we plotted the methylation levels on chromosome 14. From inside to outside: methylation levels, ideogram and gene labels. Hypermethylation (red dots and green background) and hypomethylation events (green dots and red background) in patients with lung cancer versus non-tumoral samples. (Continued) Oncotarget4398 www.impactjournals.com/oncotarget DLK1-DIO3 methylation profile by histological subtypes Due to its specific global hypomethylation in lung cancer, we have evaluated the methylation of the DLK1-DIO3 cluster in the two main histological subtypes of NSCLC, squamous cell carcinoma (SCC) and adenocarcinoma, as compared to nontumoral tissue. The DLK1-DIO3 imprinted cluster hypomethylation was consistently evident in both Figure 1: (Continued) Methylation profile of the DLK1-DIO3 cluster in lung cancer. B. Observed methylation changes (log2 ratio). Relative levels of methylation in patients with lung cancer relative to the control group are represented in blue bars, whereas methylation levels of COPD patients with respect to the control group are represented by red bars. A grey background or asterisks represent statistically significant differences (adjusted p-value < 0.05) of methylation levels with respect to the control group. NA: Not available. Oncotarget4399 www.impactjournals.com/oncotarget Table 1: Statistical differences between miRNAs and SNORDs in human tumor samples compared to non-tumoral tissue Genes Adjusted p-value miRNAs miR-770 2.51232E-14 miR-493 5.81898E-11 miR-337 7.41923E-06 miR-665 1.16877E-08 miR-431 7.32285E-13 miR-433 2.10219E-11 miR-127 1.32018E-08 miR-432 2.10017E-08 miR-136 6.70757E-09 miR-370 6.16842E-11 miR-379 2.90118E-17 miR-411 3.10121E-17 miR-299 7.69421E-17 miR-1197 1.37059E-15 miR-323A 1.73804E-15 miR-758 1.45179E-14 miR-380 1.13522E-14 miR-329-1 6.75807E-14 miR-329-2 1.64162E-13 miR-494 3.65725E-12 miR-1193 3.65725E-12 miR-543 9.62007E-09 miR-495 6.14013E-16 miR-376C 2.39664E-11 miR-654 3.35623E-12 miR-376A1 8.48877E-13 miR-300 1.65112E-14 miR-1185-1 1.86087E-11 miR-1185-2 6.87418E-14 miR-381HG 1.17148E-14 miR-381 7.43382E-11 miR-487B 3.24504E-12 miR-539 4.29749E-14 miR-889 2.92232E-16 miR-544A 1.15212E-12 miR-655 1.48893E-07 miR-487A 8.934E-13 miR-382 1.90249E-14 miR-485 1.90249E-14 miR-323B 1.05772E-13 miR-154 3.35275E-17 miR-496 2.18282E-14 miR-377 1.07483E-09 Genes Adjusted p-value miR-541 3.62941E-15 miR-409 1.59813E-13 miR-412 4.93041E-13 miR-369 2.73461E-12 miR-410 9.88918E-12 miR-656 5.71628E-12 SNORDs SNORD113-1 2.85848E-13 SNORD113-2 1.06116E-08 SNORD113-3 2.81624E-08 SNORD113-4 8.3473E-11 SNORD113-5 1.37034E-10 SNORD113-6 5.34803E-09 SNORD113-7 0.003019259 SNORD113-8 1.63653E-07 SNORD113-9 1.80412E-09 SNORD114-1 5.89536E-14 SNORD114-3 0.001522221 SNORD114-4 7.88509E-06 SNORD114-5 1.21425E-08 SNORD114-6 4.64279E-06 SNORD114-7 9.64877E-06 SNORD114-9 0.000312165 SNORD114-10 6.11146E-06 SNORD114-11 4.40691E-09 SNORD114-12 1.6391E-09 SNORD114-13 3.02053E-09 SNORD114-14 1.02738E-07 SNORD114-15 1.36835E-07 SNORD114-16 1.36239E-10 SNORD114-17 8.38123E-14 SNORD114-18 6.61886E-14 SNORD114-19 6.61886E-14 SNORD114-20 0.009700679 SNORD114-21 1.49467E-07 SNORD114-22 1.06972E-08 SNORD114-23 1.45598E-08 SNORD114-24 8.99314E-10 SNORD114-25 4.74915E-08 SNORD114-26 2.01146E-06 SNORD114-27 6.35641E-09 SNORD114-28 2.50354E-07 SNORD114-29 9.6575E-05 SNORD114-30 0.071177541 SNORD114-31 9.91748E-07 SNORD114-31 9.91748E-07 SNORD114-31 9.91748E-07 (Continued ) Oncotarget4400 www.impactjournals.com/oncotarget histological subtypes with respect to non-tumoral tissue (Supplementary Table S1). We found that SCC samples showed significantly higher hypomethylation levels in the SNORD cluster than in lung adenocarcinoma (Figure 5A). As with SNORD114-30 for general lung cancer, we did not find any statistically significant differences for lung adenocarcinoma or SCC (adjusted p-value=0.494 and 0.095, respectively). SNORD113-7 and SNORD114-3 were only statistically significant in the SCC group (log2 ratio SCC+/ non-tumoral tissue = -0.309; adjusted p-value= 0.025 and log2 ratio SCC+/ non-tumoral tissue= -0.234; adjusted p-value = 0.031, respectively) (Supplementary Table S1). MicroRNA clusters were also notably more hypomethylated in SCC (Figure 5B). RTL1, MEG8 and LINC00524 were hypomethylated in both histological subgroups. Moreover, DLK1 and MEG9 were significantly hypomethylated in patients with SCC (p<0.001 and p=0.01, respectively), while in lung adenocarcinoma, DLK1 was very similar to the control group (p=0.127). DIO3 was significantly hypermethylated in patients with lung adenocarcinoma (p<0.001) (Figure 5C). Transcriptional mapping of the DLK1-DIO3 cluster To analyze genomic features associated with different mechanisms of the transcriptional regulation of the DLK1-DIO3 cluster, we used the Wash U Epigenome Browser to display the epigenomic mapping of the DLK1-DIO3 cluster (14q32) to the human genome reference (hg19; chr14: 101.140.120–102.044.779 genomic coordinates) (Figure 6A). We observed the distribution of several transcriptional mechanisms (direct regulators and structural determinants) along the DLK1-DIO3 cluster (Figure 6B). Multiple enhancers are distributed in the length of the cluster (yellow arrows), with the exception of the region located between 101380- 10530Kb on chromosome 14. The SNORD113 and SNORD114 families and a miRNA cluster are located in this region without enhancers. In this same region, we found few effects of polycomb protein-mediated epigenetic regulation (grey arrows). In addition, we identified a heterochromatin rich domain situated approximately 101277Kb distal to the 5’ region (purple arrows). Therefore, following with MEG9 lncRNA (101536Kb), we found a highly condensed region before it reached the 5′ region of the DIO3 gene. Interestingly, the 5’ flanking region of the MEG3 lncRNA (~101292 kb) showed activity in transcription initiation, where there is an active transcriptional start site (TSS) marked by a red arrow. No other region of the DLK1-DIO3 cluster showed active TSS. From the active TSS, we found an approximately 240Kb a domain with strong transcription (green arrows) (Figure 6B). However, we identified Figure 2: Validation of the DNA methylation status of the DLK1-DIO3 cluster using TCGA data. The plot shows the methylation analysis of the DLK1-DIO3 cluster in lung cancer versus non-tumoral tissue (100 vs. 32 samples) of current and former smokers visualized by UCSC browser. Zoomed-in views of the SNORD114-9, miR-889 and DIO3 regions are highlighted in orange squares. A. Chromosome 14-scale. B. CpG sites included in the Illumina Infinium Human Methylation 450 BeadChip. C. Differential DNA methylation in lung cancer versus non-tumoral tissue of current and former smokers. Bars represent hypermethylated (red) and hypomethylated (blue) probes in lung cancer tissue with respect to normal tissue. D. ß-value differences between lung cancer and non-tumoral tissue. E. Gene names of the members of the DLK1-DIO3 cluster, and F. the position of the CpG islands present in the DLK1-DIO3 cluster (green). Oncotarget4401 www.impactjournals.com/oncotarget several CpG islands located near genes throughout this region, including lncRNAs, miRNAs, SNORDs and pseudogenes (crimson lines) (Figure 6C). However, the CpG sites were distributed along the DLK1-DIO3 cluster (white lines) (Figure 6D). The GC-content was reduced by 30% from MEG3 to MEG9 (Figure 6E). Experimentally validated target interactions of the DLK1-DIO3 miRNA cluster To study the functional relevance of the DLK1- DIO3 miRNA cluster in the regulation of validated mRNAs as a potential tumor suppressor, we used the miRWalk algorithm [27]. A total of 538 genes were found to be validated targets of the miRNAs included in this cluster. These targets were located on all chromosomes except the Y chromosome (Figure 7). Several miRNAs were complementary with more than one validated gene, and some targets were regulated by more than one miRNA. We classified validated targets by gene ontology (GO) molecular function and biological processes using the PANTHER program. The main GO molecular functions were binding (GO:0005488) (38.9%), catalytic activity (GO:0003824) (27.4%), nucleic acid binding transcription factor activity (GO:0001071) (13.5%), and receptor activity (GO:0004872) (10.4%). The primary binding types were protein binding (GO:0005515) and nucleic acid binding (GO:0003676). In the case of catalytic activity, we primarily found activities for hydrolase (GO:0016787), transferase (GO:0016740) and enzyme regulators (GO:0007154). Validated targets displayed six main biological processes: metabolic processes (GO:0008152) (21.2%), such as cell communication (GO:0030234) and cell cycle (GO:0007049); cellular process (GO:0009987) (20.2%), including system development (GO:0048731), anatomical structure and morphogenesis (GO:0009653) and death (GO:0016265); developmental processes (GO:0032502) (12.4%); response to stimulus (GO:0050896) (5.6%); and apoptotic processes (GO:0006915) (5.3%) (Supplementary Figure S1). Finally, we found that 26 out of 538 validated genes (AXIN2, BRCA1, CCNB1, CCND2, CDKN1A, CDKN2A, CSNK2A1, FOXO1, GRB2, HDAC6, NFKB1, NTRK3, PPP2R4, PPP2R1B, PPP2R2A, PTEN, RARA, RASA1, RB1, SOCS3, TGFB1, TGFB2, TGFBR2, TNF, TP53, and TP73) were targets of miRNAs included in the DLK1-DIO3 cluster, which have been proposed to have a functional role as tumor suppressors (Supplementary Table S2). Figure 3: mRNA expression levels of some DLK1-DIO3 cluster components. Expression of some DLK1-DIO3 cluster components, which was validated by real-time PCR. mRNA expression levels were determined in tumor samples and paired normal lung tissue from lung cancer patients. Data derived from RT-qPCR are presented as the log2 2-ΔΔCt values. Oncotarget4402 www.impactjournals.com/oncotarget DISCUSSION In the present study we firstly identified characteristic methylation pattern of the DLK1-DIO3 cluster in human lung cancer, as compared, to lung parenchyma. Specifically, we identified that all components of the aforementioned cluster are hypomethylated, with the exception of DIO3 that is hypermethylated. Importantly this specific methylation signature of the imprinted cluster is restricted to smoking induced lung cancer since it is absent in lung tumors from never smokers. In addition, this locus was more aberrantly regulated in SCC, typically exposed to heavier smoking history, than in lung adenocarcinoma. In a murine lung carcinoma model with mutated KRAS, it has been observed that the DLK1-DIO3 miRNAs cluster on chromosome on 12qF1 was substantially overexpressed, suggesting a role as an oncogenic driver of lung cancer in these circumstances [26]. Those findings are consistent with our results, particularly considering KRAS is the most frequently mutated proto-oncogenes among smokers with lung carcinoma [28, 29]. Imprinted genes are particularly more susceptible to epigenetic regulatory changes due to having one specifically silenced allele [6]. Therefore, monoallelic changes in methylation status may have deleterious consequences. The DLK1-DIO3 cluster includes protein-coding genes, lncRNAs, miRNAs, SNORDs and pseudogenes [21]. DLK1, RTL1 and DIO3 are expressed only from the paternal allele. However, large and small noncoding RNA genes and pseudogenes come from the maternal allele [30]. Our data showed that the paternally derived DLK1 and RTL1 genes are hypomethylated and that the DIO3 gene is hypermethylated. DLK1 has been reported aberrantly expressed in various types of tumors, such as hepatocellular carcinoma, glioma, renal cell carcinoma and lung cancer [31-34]. In addition, Li et al. recently identified in an in vitro lung model that overexpression of DLK1 enhanced extracellular matrix invasion by Notch signaling in a dependent and independent manner [23]. In the case of the RTL1 gene, it has been proposed to act as an oncogene in hepatocarcinoma. In vivo RTL1 overexpression in the liver contributed to tumor growth [24]. Accordingly, the hypomethylation of these genes can play an important role in the pathogenesis of lung cancer. The hypermethylation of DIO3 has been reported in hematologic neoplasms [35]. We also identified higher levels of global DNA methylation in lung cancer versus normal lung tissue. Moreover, the MEG3 lncRNA has been described as a tumor suppressor; MEG3 levels were significantly downregulated in patients with NSCLC [36]. However, our study showed that the Figure 4: Association between smoking behavior and methylation profile of the DLK1-DIO3 cluster. A. DLK1-DIO3 cluster methylation level differences between non-tumoral and tumoral tissue in non-smoking-related lung cancer patients. (Continued) Oncotarget4403 www.impactjournals.com/oncotarget Figure 4: (Continued) Association between smoking behavior and methylation profile of the DLK1-DIO3 cluster. B. Validation of the DNA methylation status of DIO3, miR-889 and SNORD114-9 by 454 bisulfite sequencing using pools of 10 non-tumoral and tumoral tissue samples in non-smoking-related lung cancer patients. Each column shows the methylation status of an individual CpG and the rows (and also the white numbers) indicate the amount of reads. The red arrowheads highlight the position of the CpG represented on the Illumina Infinium Human Methylation 450 BeadChip. Methylated CpGs are displayed in purple, unmethylated CpGs in orange and gaps in white. C. DLK1-DIO3 cluster methylation level differences between normal tissue and tumoral tissue in smoking-related lung cancer patients, including current and former smokers. *: adjusted p-value<0.001. β-value: sum of methylated and unmethylated probe intensities. Oncotarget4410 www.impactjournals.com/oncotarget 46. Haller F, von Heydebreck A, Zhang JD, Gunawan B, Langer C, Ramadori G, Wiemann S, Sahin O. Localization- and mutation-dependent microRNA (miRNA) expression signatures in gastrointestinal stromal tumours (GISTs), with a cluster of co-expressed miRNAs located at 14q32.31. J Pathol. 2010; 220:71-86. 47. Fu G, Ye G, Nadeem L, Ji L, Manchanda T, Wang Y, Zhao Y, Qiao J, Wang YL, Lye S, Yang BB, Peng C. MicroRNA- 376c impairs transforming growth factor-beta and nodal signaling to promote trophoblast cell proliferation and invasion. Hypertension. 2013; 61:864-872. 48. Ye G, Fu G, Cui S, Zhao S, Bernaudo S, Bai Y, Ding Y, Zhang Y, Yang BB, Peng C. MicroRNA 376c enhances ovarian cancer cell survival by targeting activin receptorlike kinase 7: implications for chemoresistance. J Cell Sci. 2011; 124:359-368. 49. Luk JM, Burchard J, Zhang C, Liu AM, Wong KF, Shek FH, Lee NP, Fan ST, Poon RT, Ivanovska I, Philippar U, Cleary MA, Buser CA, et al. DLK1-DIO3 genomic imprinted microRNA cluster at 14q32.2 defines a stemlike subtype of hepatocellular carcinoma associated with poor survival. The Journal of biological chemistry. 2011; 286:30706-30713. 50. Nadal E, Zhong J, Lin J, Reddy RM, Ramnath N, Orringer MB, Chang AC, Beer DG, Chen G. A MicroRNA cluster at 14q32 drives aggressive lung adenocarcinoma. Clinical cancer research: an official journal of the American Association for Cancer Research. 2014; 20:3107-3117. 51. Assenov Y, Muller F, Lutsik P, Walter J, Lengauer T, Bock C. Comprehensive analysis of DNA methylation data with RnBeads. Nature methods. 2014; 11:1138-1140. 52. Maksimovic J, Gordon L, Oshlack A. SWAN: Subsetquantile within array normalization for illumina infinium HumanMethylation450 BeadChips. Genome Biol. 2012; 13:R44. 53. Ritchie ME, Diyagama D, Neilson J, van Laar R, Dobrovic A, Holloway A, Smyth GK. Empirical array quality weights in the analysis of microarray data. BMC Bioinformatics. 2006; 7:261. 54. Krzywinski M, Schein J, Birol I, Connors J, Gascoyne R, Horsman D, Jones SJ, Marra MA. Circos: an information aesthetic for comparative genomics. Genome Res. 2009; 19:1639-1645. 55. Zhou X, Li D, Lowdon RF, Costello JF, Wang T. methylC Track: visual integration of single-base resolution DNA methylation data on the WashU EpiGenome Browser. Bioinformatics. 2014; 30:2206-2207. 56. Mi H, Muruganujan A, Thomas PD. PANTHER in 2013: modeling the evolution of gene function, and other gene attributes, in the context of phylogenetic trees. Nucleic Acids Res. 2013; 41:D377-386. 57. Geyh S, Rodriguez-Paredes M, Jager P, Khandanpour C, Cadeddu RP, Gutekunst J, Wilk CM, Fenk R, Zilkens C, Hermsen D, Germing U, Kobbe G, Lyko F, et al. Functional inhibition of mesenchymal stromal cells in acute myeloid leukemia. Leukemia. 2016; 30:683-691. 58. Gronniger E, Weber B, Heil O, Peters N, Stab F, Wenck H, Korn B, Winnefeld M, Lyko F. Aging and chronic sun exposure cause distinct epigenetic changes in human skin. PLoS genetics. 2010; 6:e1000971. 59. Molina-Pinelo S, Gutierrez G, Pastor MD, Hergueta M, Moreno-Bueno G, Garcia-Carbonero R, Nogal A, Suarez R, Salinas A, Pozo-Rodriguez F, Lopez-Rios F, Agullo-Ortuno MT, Ferrer I, et al. MicroRNA-dependent regulation of transcription in non-small cell lung cancer. PloS one. 2014; 9:e90524. 60. Molina-Pinelo S, Pastor MD, Suarez R, Romero-Romero B, Gonzalez De la Pena M, Salinas A, Garcia-Carbonero R, De Miguel MJ, Rodriguez-Panadero F, Carnero A, Paz-Ares L. MicroRNA clusters: dysregulation in lung adenocarcinoma and COPD. Eur Respir J. 2014; 43:1740-1749.