Title DNA methylation of microRNA genes in multiple myeloma

DNA methylation is one of the heritable epigenetic modifications, leading to repressed gene expressions and consequent phenotypic alterations without changing the DNA sequence. MicroRNA (miRNA) is a novel class of short non-coding RNA molecules regulating a wide range of cellular functions through translational repression of their target genes. Recently, epigenetic dysregulation of tumor suppressor miRNA genes by promoter DNA methylation has been implicated in human cancers, including multiple myeloma (MM). This article presents a brief overview of the pathogenesis of MM, the role of DNA methylation in cancer biology, methods of DNA methylation analysis, miRNA biology, dysregulation of miRNAs in MM, and summaries the current data on the role of DNA methylation of tumor suppressive miRNAs in MM.

non-hyperdiploid and hyperdiploid MM [8][9][10] (Figure 1). It has also been found that this ploidy dichotomy can also be detected even in MGUS, the precursor of symptomatic MM, and hence an early event of myelomagenesis [10,11].
In contrast to the non-hyperdiploid MM, hyperdiploid MM constitutes another half of all MGUS and MM, due to trisomies of odd-numbered chromosomes including 3, 5, 7, 9, 11, 5 15, 19 or 21, resulting in a chromosome number between 46/47 to 75, as measured by conventional karyotyping; and/or a DNA index between 1.05 to 1. 75. In contrast to other odd-numbered chromosomes frequently involved in hyperdiplody, chromosome 13 is often deleted instead [20,21]. Moreover, trisomy of chromosome 11, to which CCND1 is localized, has been shown to result in direct upregulation of CCND1 [22,23]. While the underlying mechanism leading to the aforementioned dichotomy of MM remains to be elucidated, it is noteworthy that, clinically, hyperdiploid MM patients are associated with better prognosis and treatment outcomes than the non-hyperdiploid MM patients [24][25][26].
During disease progression, secondary translocations and other genetic aberrations, including deletion of the short arm of chromosome 17 [del(17p)] and mutations of RAS genes, etc, are involved [27]. Unlike primary translocations, which involve juxtaposition of the strong immunoglobulin heavy chain gene enhancer locus to a partner oncogene, the mechanisms of secondary translocations are less well-defined but appears unrelated to the error-prone B cell-specific DNA modification events. For instance, at the time of disease progression, about 15% of MM patients carry secondary immunoglobulin heavy chain translocation involving v-myc myelocytomatosis viral oncogene homolog (avian) (c-MYC) (8q24), which confers proliferative advantage to MM cells [27,28].
Based on fluorescence in situ hybridization (FISH) analysis, del (13) is detected in 20% to 50% of MGUS, and approximately 50% of MM. Notably, 90% of del(13) is characterized by monosomy 13, and interstitial deletion of 13q14 occurs in the remaining 10% of cases [20,21]. Despite that del (13) was once believed to impart poor prognosis, recently, the 6 prognostic impact of del (13) has been shown to be mediated by its strong association with unfavorable risk factor of t(4;14) [29,30].
By interphase FISH analysis, del(17p), which is the locus for tumor suppressor protein TP53, is generally found in less than 10% of MM patients at diagnosis. However, presence of del(17p) at diagnosis is a powerful negative prognostic factor for MM [31,32]. A recent study of MM patients uniformly receiving bortezomib-based induction therapy prior to autologous stem cell transplantation further confirmed that del(17p) is associated with an inferior event-free survival (median time: 14 vs. 36 months) and overall survival (4-year OS: 50% vs. 79%) as compared with those without del(17p). Therefore, the adverse impact of del(17p) appears not abolished by the use of targeted therapy [33].
RAS mutations, predominantly K-and N-RAS at codon 12, 13 and 61, but not H-RAS, is present in more than half of MM at diagnosis but not in MGUS, suggesting that the RAS mutations is at least a marker of the transition from MGUS to MM [34][35][36][37].
Last but not least, the bone marrow microenvironment is very important in the pathogenesis of MM for homing of MM plasma cell to the bone marrow and secretion of growth-stimulating cytokines to the MM plasma cells. The homing of MM plasma cell is a chemotaxis mechanism mediated by the bone marrow stromal cells secreted chemokine (C-X-C motif) ligand 12 or stromal cell-derived factor 1, which binds to its specific receptor, chemokine (C-X-C motif) receptor 4, expressed on the MM plasma cell. Moreover, upon cell-to-cell interactions between the bone marrow stromal cells and the MM plasma cells, a 7 multitude of cytokines are secreted, and hence favor the proliferation and survival of MM plasma cells by autocrine and paracrine signaling [e.g. interleukin 6 (IL6), insulin-like growth factor 1, tumor necrosis factor alpha and nuclear factor of kappa light polypeptide gene enhancer in B-cells (NFκB)], angiogenesis [e.g. vascular endothelial growth factor and basic fibroblast growth factor] and osteolysis [e.g. tumor necrosis factor (ligand) superfamily, member 11, tumor necrosis factor receptor superfamily, member 11a and tumor necrosis factor receptor superfamily, member 11b] [27,38,39].

DNA methylation and cancer
DNA methylation refers to catalytic addition of a methyl group (-CH) to carbon 5 position of a cytosine ring in a CpG dinucleotide [40][41][42][43][44]. CpG dinucleotide cluster in any genomic region of over 200 bp in length with a high GC content of more than 50% and observed/expected CpG ratio larger than 0.60 is known as a CpG island [45,46]. In human, CpG island is associated with at least 50% of gene promoters. Methylation of a promoter-associated CpG island will lead to recruitment of histone methyltransferase, methyl-CpG-binding domain (MBD) protein and histone deacetylase, resulting in formation of a compact heterchromatin configuration, which precludes the binding of transcription factor complex, and hence silencing of the associated gene [46,47].
In normal cells, a majority of genes with promoter-associated CpG islands are usually unmethylated, associated with a euchromatin configuration, and hence are generally transcriptionally ready or active ( Figure 2). However, a fraction of genes with 8 promoter-associated CpG islands are methylated, and hence silenced in normal cells including genetic imprinting and X-chromosome inactivation [48,49].
On the other hand, in cancer cells, genes with promoter-associated CpG islands are aberrantly methylated in a tumor-specific manner, leading to gene silencing. In particular, hypermethylation of promoter-associated CpG islands of tumor suppressor genes, resulting in decreased or loss of gene expressions, and hence loss of tumor suppressor functions, has been implicated in carcinogenesis [40,41,43,44,[50][51][52] (Figure 2). Furthermore, in cancers, hypermethylation of the tumor suppressor genes may serve as a second hit, in addition to deletion or mutation of the other allele, thereby fulfilling the Knudson's two-hit hypothesis [53].
In MM, by genome-wide or gene-specific approaches, aberrant DNA methylation has been found to mediate the loss of a number of protein-coding tumor suppressor genes regulating cell cycle progression, cell signaling or apoptosis, including cyclin-dependent kinase inhibitor 2A (CDKN2A), cyclin-dependent kinase inhibitor 2B (CDKN2B), death-associated protein kinase, secreted frizzled-related protein 2 and suppressor of cytokine signaling 1 (SOCS1), etc [54,55].

Methods of DNA methylation analysis
Over the years, techniques of DNA methylation analysis have evolved from qualitative to quantitative in fashion, and from locus-specific to genome-wide in scale [56][57][58] (Table 1).
Bisulfite conversion, which chemically deaminates or modifies unmethylated cytosine to 9 uracil, and hence translating an epigenetic variation (methylated or unmethylated) into a genetic difference (C or U), is an important procedure fundamental to most of the later methods [59,60].

(I) Candidate gene-specific methods
Before the integration of bisulfite conversion into DNA methylation research, earlier techniques mostly depend on enzymatic digestion and high-performance liquid chromatography (HPLC). For the methylation-sensitive restriction enzyme-based analysis, for example, DNA methylation pattern of CpG dinucleotides embedded in a CCGG sequence can be detected by the use of isoschizomer pair of HpaII and MspI, by which methylated CCGG can be digested by MspI but not by HpaII, together with gel electrophoresis and Southern blotting [61]. However, the use of restriction enzyme digestion is limited by the requirement of large amount of DNA and the availability of restriction enzyme cut site at the locus-of-interest. Moreover, it is less informative about the methylation pattern over a stretch of CpG dinucleotides and prone to generate false-positive results due to incomplete digestion. In addition to enzymatic digestion, HPLC was also employed in DNA methylation analysis in the early days [62]. However, the use of HPLC is also limited by the requirement of large amount of DNA and the need of skillful and tedious operation. MSP is currently the most popular technique used in studying DNA methylation of locus-specific CpG sites because of its specificity and simplicity [64]. DNA methylation status of any given CpG site is revealed by PCR amplification of bisulfite-converted DNA with two sets of PCR primers, one specific to the methylated sequence and the other to the unmethylated sequence. With validated specific primers and PCR conditions, the methylation status in a large number of samples can easily be obtained. However, it is not a quantitative method.
COBRA is highly similar to one of the two classic methods for the use of restriction enzyme, however, with the incorporation of bisulfite conversion, it becomes a high-throughput and quantitative technique [65]. Upon bisulfite conversion, unmethylated BstUI recognition sites CGCG will be converted to TGTG, whereas methylated BstUI sites 11 remain unchanged. Followed by BstUI digestion and gel electrophoresis, methylation of a locus-of-interest can be quantified by [100% X intensity of (digested fragments/ both digested and undigested fragments)].
Ms-SNuPE is a quantitative DNA methylation analysis method derived from primer extension technique [66]. In brief, primer extension is performed on bisulfite-treated locus-of-interest with 32 P-labeled dCTP or dTTP, which enable differentiation and quantification of the methylated or unmethylated template. However, the use of radioactive isotopes hinders the popularity of this technique.
MethyLight, also known as quantitative MSP, enables simultaneous detection and quantification of bisulfite-treated methylated and unmethylated templates by two specific TaqMan probes labeled with different fluorophores [67]. By the use of real-time PCR, MethyLight is regarded as a high-throughput, sensitive and quantitative method in DNA methylation research.
Pyrosequencing, which originally designed to study single-nucleotide polymorphism by indirect detection of pyrophosphate (PPi) released during DNA synthesis, has also been applied to detect the C and T difference generated by bisulfite conversion [68,69]. During a primer extension process on a bisulfite-treated template, PPi is released in an equimolar fashion according to the number of incorporated nucleotides, resulting in a proportional conversion of PPi to ATP by sulfurylase, and hence a quantifiable firefly luciferase signal 12 driven by the ATP. However, such a high-throughput, accurate and quantitative method is limited by the length of individual read, which is only about 60 to 100 bp.
MassARRAY, a technique involving a combination of bisulfite conversion, in vitro transcription, RNA digestion and MALDI-TOF mass spectrometry, is another high-throughput quantitative DNA methylation analysis method [70]. In brief, bisulfite-treated locus-of-interest is amplified with an in vitro transcription tag, which allows later in vitro transcription. The transcripts will then be digested into fragments without affecting any of the original CpG sites. Based on mass difference arise from methylated (resulting G) and unmethylated (resulting A) on the fragments, quantification of the methylated or unmethylated fragment is enabled. However, this technology requires sophisticated operation.

(II) Genome-wide methods
Recently, coupled with bisulfite conversion, methylation-sensitive restriction enzyme digestion or methylation-sensitive antibody purification, genome-wide analysis of DNA methylation is made possible with different kinds of DNA microarrays and high-throughput sequencing methods [57,[71][72][73].
Bisulfite-converted DNA, for instance, can be subjected to Infinium Methylation Assay (Illumina), which allows quantitative analysis of > 485,000 specific CpG dinucleotides per sample. By methylation-specific single-base primer extension, specific fluorescence-labeled nucleotides will be incorporated, and hence a ratio of different fluorescent signals indicating 13 the methylation status [74,75]. Other array platforms include Affymetrix, Agilent and NimbleGen. Alternatively, bisulfite-converted DNA libraries can be generated from sonication or restriction enzymes prior to high-throughput sequencing, resulting in whole-genome bisulfite sequencing (WGBS) or reduced representation bisulfite sequencing (RRBS), which is able to generate DNA methylome with single-base resolution [76,77].
Moreover, methylation-sensitive restriction enzyme may be used to enrich methylated or unmethylated DNA for different kinds of tiling arrays and high-throughput sequencings.
For instance, BstUI or HpaII digestion (cleaves unmethylated DNA) will lead to enrichment of methylated sequences, whereas MspI or McrBC digestion (cleaves methylated DNA) will result in enrichment of unmethylated sequences, which are followed by array or sequencing profiling [78,79].
Alternatively, with the development of methylation-sensitive antibodies, such as MECP2 (methyl-CpG-binding protein 2), MBD1 and MBD2, which binds to methylated CpG sites, immunoprecipitation of methylated sequences prior to DNA microarrays or high-throughput sequencing is enabled and collectively known as MeDIP (methylated DNA immunoprecipitation)-chip, when it is analyzed by DNA microarray, or MeDIP-seq, when it is analyzed by high-throughput sequencing method [80,81].

MicroRNA (miRNA)
History and biogenesis 14 miRNA is a class of short non-coding RNA molecules of 20 to 30 nucleotides (nts) in length [82]. miRNAs inhibit the translation of their own target genes via binding of the miRNA seed region (i.e. the 2nd to 7th nts from 5' to 3' of the mature miRNA) to the three prime untranslated region (3'UTR) of the target gene, and hence involve in the regulation of various cellular activities, including development, differentiation, proliferation, and apoptosis [83][84][85][86] (Figure 3). Most miRNA genes are associated with RNA polymerase II promoter and are generally first transcribed into primary (pri-) miRNA (>100 nts) by RNA polymerase II in the nucleus [92]. A pri-miRNA transcript is first stabilized by 5'capping and 3'polyadenylation, and then further processed into precursor (pre-) miRNA by RNase III Drosha and its co-factor Pasha [93,94]. A pre-miRNA (60 to 80 nts) forms a hairpin or stem-loop structure, followed by export into the cytoplasm through Ran-GTP-dependent exportin 5 [95,96]. In the cytoplasm, the pre-miRNA is further processed by RNase III Dicer into mature miRNA duplex (22 to 25 nts), which will then be loaded into a RNA-induced silencing complex (RISC) [94]. The functional mature miRNA strand of the duplex is retained in the RISC for recognizing the mRNA target through sequence complementarity between the miRNA seed region and the 15 3'UTR of the target gene. Subsequently, the target gene is inhibited by either translational inhibition or mRNA degradation mediated by the miRNA-associated RISC [97,98].

miRNA, DNA methylation and Knudson's hypothesis
Mechanistically, miRNAs play a role in the regulation of DNA methylation in cancers. In breast and cervical cancers, CDKN2A and SOCS1 are important tumor suppressor genes inactivated by DNA hypermethylation. [103,104]. On the other hand, miR-24 and 16 miR-155 are oncogenic miRNAs over-expressed in breast and cervical cancers [105,106].
Furthermore, miR-24 and miR-155 have been shown to target the 3'UTR of, and hence repress CDKN2A and SOCS1 respectively. [106,107]. Therefore, in addition to gene hypermethylation, tumor suppressor genes can be translationally repressed by oncogenic miRNAs, suggesting that the Knudson's hypothesis can potentially be fulfilled by a complex co-operation of gene alterations with one allele inactivated by gene deletion, mutation or hypermethylation, and the other allele by miRNA targeting.
Thus, these data suggested that miRNAs play multifaceted role in carcinogenesis. Later part of this article focuses on the role of DNA methylation of tumor suppressor miRNAs in MM.
Furthermore, of these dysregulated miRNAs, some were found to be involved in the regulation of cell cycle, proliferation and apoptosis. For examples, miR-21 and miR-  were identified as oncogenic miRNAs, leading to enhanced survival and reduced apoptosis of MM [110,118]. In particular, upregulation of miR-21, a downstream target of activator of transcription 3 (STAT3), was found to potentiate the proliferative IL6-mediated signal transducer and STAT3 signaling in MM. In contrast, miR-15a and miR-16-1 were found to be tumor suppressor miRNAs, resulting in increase of apoptosis and suppression of NFκB 17 pathway in MM cells [111,119]. Moreover, these studies showed that some known tumor suppressor miRNAs, such as let-7, miR-29 and miR-193, are downregulated in MM.
Downregulation of miRNA expression in cancers may be mediated by various mechanisms, ranging from epigenetic inactivation, gene mutation or copy number loss to defective miRNA biogenesis or post-transcriptional processing [120]. Of these, DNA methylation is associated with repression of miRNAs possessing promoter-associated CpG islands [121]. Furthermore, the expressions and functions of these tumor suppressor miRNAs can be reversed and restored by DNA hypomethylation treatment [122]. Therefore a better understanding of epigenetic inactivation of tumor suppressor miRNA genes is essential for the biology and treatment in human cancers including MM. Recently, the following studies described the role of DNA methylation of tumor suppressor miRNA genes, including miR-34a, miR-34b/c, miR-194-2-192, miR-203 and miR-124-1, in MM.

miR-34a
The tumor suppressor protein TP53 plays a central role in the tumor suppression network, in response to carcinogenic cellular stress and DNA damage, through the induction of apoptosis, cell cycle arrest and senescence [123]. Deletion of the short arm of chromosome 17, to which TP53 gene is localized, confers an adverse impact on event-free and overall survival of MM patients [29,124]. However, homozygous deletion or mutation of the TP53 gene is rarely found in MM patients [32,125]. Therefore, it was hypothesized that 18 the dysregulation of the TP53-mediated tumor suppression may be due to inactivation of other tumor suppressive components along the TP53 pathway in MM.
With the presence of promoter-associated CpG island at each of the miR-34a (1p36) and miR-34b/c (11q23) promoters, frequent DNA hypermethylation of the miR-34s gene, leading to silencing of the miR-34s, and hence upregulation of the miR-34s target genes has been found in a wide range of solid cancers, including bladder, breast, colon, lung, melanoma, neuroblastoma, prostate, and ovarian cancer, etc [126,127,129,130,[140][141][142][143].  [144,145]. Both of these studies showed that the promoter-associated CpG island of 19 the miR-34a was unmethylated in normal controls but aberrantly methylated in 50% of the hematological cancer cell lines, including human myeloma cell lines (HMCLs). Treatment with 5-aza-2'-deoxycytidine led to demethylation of the miR-34a promoter and consequent re-expression of the pri-miR-34a transcript in cells homozygously methylated for the miR-34a. Among primary samples at diagnosis, the miR-34a promoter was preferentially methylated in 18.8% NHL (p=0.018), 5.5% MM, 4.0% CLL and 2.2% MDS and none of ALL, AML and CML. Furthermore, in MM, with paired primary samples of at diagnosis and at relapse/progression, it was also shown that methylation of the miR-34a promoter remained infrequent even at the time of disease relapse/progression. Therefore, in contrast to the frequent methylation of miR-34a in epithelial cancers [142,146], methylation of the miR-34a promoter appears unimportant in MM pathogenesis and progression.

miR-34b/c
In contrast to miR-34a, methylation of the miR-34b/c was implicated in the progression of MM [147]. In a recent study, the promoter-associated CpG island of the miR-34b/c was shown to be unmethylated in normal controls but aberrantly methylated in 75% of the

miR-203
Epigenetic inactivation of the tumor suppressor miRNA miR-203, localized to 14q32, was reported in CML, hepatocellular carcinoma and a wide range of hematological malignancies [145,[149][150][151]. While juxtaposition of the 14q32 immunoglobulin heavy chain enhancer to an oncogene partner occur in approximately 50% of MGUS and SMM, > 75% of MM, and > 80% of PCL, leading to upregulation of oncogenes, such as CCND1, CCND3, FGFR3, MMSET, and MAF [12,13], double-stranded DNA breaks inherent with the translocation may result in DNA loss [152], and hence potential loss of tumor suppressor gene or miRNA. Wong Moreover, methylation of these tumor suppressor miRNAs can be reversed by hypomethylation treatment, leading to restoration of corresponding expression and tumor suppressor function of these miRNAs. Epigenetic therapy has recently emerged as a state-of-the-art strategy in cancer treatment [159,160]. For instance, pharmacological grade DNA methyltransferase inhibitors have been approved for the treatment of myelodysplastic syndrome [161]. Therefore, these findings on methylation of tumor suppressor miRNAs may provide a foundation for the use of epigenetic drugs in the treatment of MM. In addition, 24 these data also suggest the potential use of tumor suppressor miRNA mimics as a cancer therapy in tumors lacking certain critical tumor suppressor miRNAs [162].
Last but not least, these data highlight the importance of methylation of tumor suppressor miRNAs in MM with respect to the disease pathogenesis, diagnosis and therapy.
Therefore, future genome-wide analysis of DNA methylation of miRNAs in MM will allow identification of novel miRNAs important in myelomagenesis.