Background
MicroRNAs (miRNAs) are small non-coding RNAs that participate in the control of many cellular processes such as stress response, cell differentiation, cell-cycle regulation, stem cell biology, apoptosis among many others [
1]. MicroRNAs exert their regulatory effect post-transcriptionally by inducing RNA degradation or translation inhibition, and their expression can be deregulated in cancer by genetic and epigenetic mechanisms [
2‐
4]. MicroRNAs can also affect gene expression of many genes by direct regulation of the epigenetic machinery. For example, microRNAs like
miR-101,
miR-205 and miR-26a regulate chromatin modifiers in cancer such as the Polycomb associated histone methyltransferase EZH2 [
2,
3]. The DNA methylation maintenance enzyme Dnmt1 is regulated in different cell- types by the
miR-126 and
miR-152, as well as the
de novo methyltransferases Dnmt3a and Dnmt3b by the
miR-29 family members
miR-29a, −
29b and
-29c [
5]. Overexpression of
miR-29a, −29b and -29c cause abnormal downregulation of the Dnmt3a and Dnmt3b, which is associated with development of lung cancer and acute myeloid leukemia [
6,
7].
DNA methylation can regulate microRNAs gene expression in cancer [
8]. In particular, repression of gene expression by DNA methylation of promoter associated CpG islands has been reported for several microRNAs in glioblastoma cells like
miR-211,
miR-204,
miR-145,
miR-137 among others [
9‐
12]. For example,
miR-145 was shown to be downregulated in glioblastoma cells and low expression of
miR-145 was found to be correlated with poor prognosis in patients [
11]. Overexpression of
miR-145 reduced cell proliferation, migration and invasion in glioblastoma cells by suppressing SOX9 and ADD3 [
13]. Thus, DNA methylation of CpG-rich microRNAs promoters in glioblastoma cells seems to be an important process for tumour development and maintenance.
CTCF is a ubiquitous, highly-conserved 11-zinc finger nuclear protein [
14,
15], which is subjected to different post-translational modifications [
16,
17] and has been implicated in a broad range of functions including higher-order chromatin organization by favoring inter- and intra-chromosomal interactions [
18‐
20]. The combinatorial usage of different zinc-fingers confers CTCF the capacity to bind complex sequences, interact with other proteins and with ncRNAs [
14,
21‐
23]. CTCF is also important to maintain, CpG-rich promoter regions of tumour suppressor genes, like
BRCA1,
retinoblastoma, and others, in an unmethylated state [
24,
25]. Importantly, DNA methylation can affect CTCF binding in part because of the presence of CpGs in the CTCF binding motif [
26]. For example, increased methylation at the promoter of the brain-derived neurotrophic factor (
BDNF) triggered the dissociation of CTCF which resulted in gene silencing [
27]. In fact 41 % of cell-type specific CTCF binding sites show differential DNA methylation [
28].
In addition, several reports have implicated CTCF in the regulation of microRNAs expression [
29].
MiR-125b expression is decreased in breast cancer, partly, through CTCF dissociation from its promoter region [
30]. In addition, ERα positive breast cancer cells overexpress
miR-375 concomitantly with promoter DNA hypermethylation and CTCF depletion [
31]. Furthermore, CTCF and pluripotency maintenance factors are depleted in the
miR-290 regulatory region in differentiated embryonic stem cells, together with increased DNA methylation and deposition of the repressive histone mark H3K27me3 [
32].
The
miR-181c is a member of the miR-181 family of microRNAs involved in the development of glioblastoma multiforme (GBM), which is one of the most frequent and malignant primary brain tumours [
33,
34].
MiR-181c is downregulated in GBM, and its expression levels correlate with tumour progression, suggesting that its epigenetic regulation could be affected [
33]. In contrast,
miR-181c is overexpressed in gastric cancer, skin basal cell carcinoma, and in osteosarcomas [
35‐
37].
Here we explored the epigenetic regulatory processes responsible for the deregulation of miR-181c in glioblastoma cells; in particular, we asked whether the nuclear factor CTCF participates in its epigenetic regulation. We first confirmed that miR-181c is differentially expressed in glioblastoma cell lines. We analyzed ChIP-seq data sets from different cell-types and identified a DNA region located in the 5′ non-coding region of the miR-181c enriched in histone marks characteristic of promoter regions. We confirmed binding of CTCF to the promoter region of miR-181c in the glioblastoma cell line U87MG and K562 cells. In contrast, CTCF does not bind the promoter region of the aggressive glioblastoma cell line T98G. Absence of CTCF correlates with gain of DNA methylation and miR-181c downregulation. Furthermore, we show that depletion of CTCF in glioblastoma cells affects the expression levels of NOTCH2 a target of miR-181c. Together, these results implicate CTCF and DNA methylation in the epigenetic regulation of miR-181c in cancer cells.
Discussion
Cancer is a multistep disease that includes many interdependent components at the cellular level [
44]. There are also molecular components that include genotypic abnormalities but more recently epigenotypic deregulation [
45]. In particular, and based on the relevance of the post-transcriptional regulatory function of microRNAs over different types of genes we studied here how epigenetic regulatory processes can dysregulate microRNAs transcription in cancer. We asked how a microRNA, the
miR-181c, involved in the regulation of brain specific genes can be epigenetically deregulated in glioblastoma cell lines, one of the more frequently occurring primary malignant brain tumours. We focused on the glioblastoma cell lines, T98G and U87MG, were the
miR-181c is downregulated in comparison to normal brain tissues. This microRNA loss of gene expression correlated with a strong gain of DNA methylation in the
miR-181c promoter region. Importantly, this aberrant DNA hypermethylation apparently interferes with the binding of the chromatin associated CTCF nuclear factor. CTCF depletion confirmed a gain of DNA methylation in U87MG cells supporting a previously reported protective role of CTCF in tumour suppressor genes [
25]. Finally, CTCF knockdown induces the upregulation of
NOTCH2 a target of
miR-181c.
Concerning the transcriptional regulation of microRNAs an important sub-group is annotated as intergenic, but others are intronic and/or exonic, either in sense or antisense orientations presenting a more complex regulatory context. Genetic disruption of microRNAs has been documented in cancer, but there are some evidences that suggest that epigenetic alterations can be one of the major mechanisms for microRNA deregulation in cancer and other diseases [
2]. There is a growing list of microRNAs that are subjected to epigenetic abnormal influence, including gain or loss of DNA methylation, histone covalent modifications, and more recently, the topological organization of the genome (see below). For example, it is well documented how members of the miR-34 family are involved in cancer through cell cycle arrest, cell invasion, apoptosis or even cancer metastasis [
2]. These microRNAs are mainly silenced by DNA methylation of their promoter regions. Concerning the role of CTCF in microRNAs, a recent report showed that the miR-125b1 is aberrantly silenced by DNA methylation in breast cancer cells [
30]. In such context, CTCF binding to the promoter region of the
miR-125b1 is disrupted and a gain in the repressive histone modification H3K9me3 and H3K27me3 is detected in cancer cells [
30]. Interestingly, alternative epigenetic silencing mechanisms exist, like the overexpression of EZH2, a key member of the Repressive Polycomb Complex, PRC2, that in addition to silence many genes, including tumour suppressor genes, can also silence different microRNAs in cancer cells [
46]. It has been documented by several research groups that EZH2 is overexpressed in different cancers, and found to repress abnormally different microRNAs, including the miR-181c in prostate and breast cancer cells [
47]. Then, based on our observation and the differential binding of CTCF to the miR-181c in different cell-types we propose that EZH2 and Polycomb proteins may be responsible for silencing the miR-181c in cell-types were the miR-181c is normally not expressed, like in the human erythroleukemic K562 cells or primary lymphocytes (Fig.
1).
An important aspect that is to a certain extent underestimated is the possibility that in glioblastoma cells CTCF is affected by mutations. Nowadays, there is a repertoire of different CTCF mutations, comprising somatic mutations, resulting in nonsense, missense, frameshift and splice site mutations [
48]. Some of these mutations have been identified in different cancer types. From a functional point of view, a large proportion of mutations are found in the zinc-fingers that are critical for CTCF binding to DNA [
48,
49]. Therefore, in glioblastoma cells and in regulatory regions as for the miR-181c, CTCF disruption can be caused by specific mutations that affect its binding to DNA. This view is further supported by a recent report in which
ctcf hemizygous knockout mice predisposes to cancer, under certain inducible conditions, promoting tumour aggressive invasion and metastatic dissemination [
50]. What is even more relevant, in the context of the present study, is the fact that CTCF haploinsufficient mice destabilize genome-wide DNA methylation patterns supporting the relationship between CTCF and DNA methylation in certain genomic regions [
50]. In the same study point mutations have been correlated with abnormal gain of DNA methylation. Therefore, CTCF is now considered as a tumour suppressor gene in human cancers and is significantly mutated gene in different types of cancers [
50,
51].
Based on the recent series of publications and given the architectural role attributed to CTCF we cannot discard, that the CTCF located in the promoter region of the miR-181c plays a structural role [
52]. Due to this possibility we analyzed the genomic distribution of CTCF, and its relationship with the three-dimensional architecture of the genome taking advantage of the newly, high resolution, genome-wide mapping of chromatin loops by
in situ Hi-C [
43].
In situ Hi-C series of experiments have reached up to 1 kb resolution. As shown in Fig.
6, the CTCF site associated with the miR-181c promoter does not seem to correspond to a loop anchor site (Fig.
6c). We believe that this is relevant, and we propose that this CTCF site is not a structural one, instead we suggest a local regulatory function, in particular, protection against DNA methylation. In addition, Lieberman Aiden and collaborators demonstrated that more than 90 % of the CTCF sites at loop anchors, at the DNA binding sequence level, are positioned in a convergent orientation [
43]. This is extremely relevant since this type of sequence convergence orientation for CTCF binding sites turns out to be an excellent predictor of chromatin loop formation. Based in such prediction we propose a model in which the miR-181c, and its adjacent gene
Nanos3, are not included in a loop and their location correspond to a genomic region between two large chromosomal loops (Fig.
6c).
In glioblastoma the Notch signaling pathway is aberrantly activated [
53]. NOTCH2 is one of the receptors of the Notch pathway and was recently shown to be important for proliferation, invasion and self-renewal of glioblastoma U87MG cells [
34]. The
NOTCH2 gene is also a post-transcriptionally target of
miR-181c and a negative correlation between
NOTCH2 gene expression and
miR-181c was found in glioblastoma samples [
34]. In the present study we observed that CTCF knockdown induces overexpression of
NOTCH2 gene in U87MG glioblastoma cells possibly as a consequence of the epigenetic silencing by DNA methylation of
miR-181c (Fig.
7a). This finding highlights the importance of CTCF as a regulator of gene expression for tumour suppressor genes. In conclusion, microRNAs are subjected to multiple levels of regulation and there are few examples of how they are regulated transcriptionally, and even fewer how they are deregulated epigenetically. Due to their critical role during animal development it is important to continue exploring how these regulatory genes are controlled by a multitude of mechanisms.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
EA-O and FR-T designed the study and wrote the manuscript. EA-O, RA-M, RP-M, EG-B, GG and KM performed the experiments. FR-T, RA-M and EA-O performed the bioinformatic analysis of the corresponding genomic region. All authors read and approved the final version of the manuscript.