Factors Regulating microRNA Expression and Function in Multiple Myeloma

Intensive research has been undertaken during the last decade to identify the implication of microRNAs (miRNAs) in the pathogenesis of multiple myeloma (MM). The expression profiling of miRNAs in MM has provided relevant information, demonstrating different patterns of miRNA expression depending on the genetic abnormalities of MM and a key role of some miRNAs regulating critical genes associated with MM pathogenesis. However, the underlying causes of abnormal expression of miRNAs in myeloma cells remain mainly elusive. The final expression of the mature miRNAs is subject to multiple regulation mechanisms, such as copy number alterations, CpG methylation or transcription factors, together with impairment in miRNA biogenesis and differences in availability of the mRNA target sequence. In this review, we summarize the available knowledge about the factors involved in the regulation of miRNA expression and functionality in MM.


Introduction
Multiple myeloma (MM) is a neoplasm characterized by the accumulation of clonal plasma cells (PC) in bone marrow, which is generally associated with the production of a monoclonal immunoglobulin and lytic bone lesions. MM accounts for 1% of all cancers and approximately 10% of all hematologic malignances. Although the incorporation of new drugs with different mechanisms of action has dramatically changed the perspective of treatment with an improvement in the median overall survival rate, it remains an incurable disease [1].
Similar to other cancer types, MM is a cytogenetically heterogeneous disease. Initial studies of conventional cytogenetics and fluorescent in situ hybridization (FISH), and more advanced studies based on microarrays and new generation sequencing (NGS), have revealed the existence of different types of genetic alterations that can be categorized into translocations, copy number abnormalities (CNAs) and point mutations [2][3][4]. The majority of chromosomal translocations affect IGH locus at 14q32, which occur in approximately 50% of MMs. Five chromosomal partners account for most of the IGH translocations in MM, 11q13, 4p16, 16q23, 6p21, and 20q11. These translocations lead to the overexpression of cyclin D1, FGFR3/MMSET, c-MAF, cyclin D3, and MAF-B, respectively. The presence miR-520a-5p, miR-518d-5p, miR-498 and miR-520g, located at 19q, has been shown to be upregulated in those MM patients with 19q gains [17]. These miRNAs target p21, involved in cell cycle regulation [29], and FOXO3, a transcription factor implicated in cell growth, proliferation, and development of cancer [30]. However, their contribution to the pathogenesis of MM is unknown. Regarding chromosomal losses, monosomy and deletions of chromosome 13 are detected in up to 50% of MM cases [2,3]. Some of the miRNAs located at this region are downregulated in primary MM samples with del(13q), like miR-15a/16 cluster and members of the miR-17-92 cluster (miR-17, miR19a and miR-20a) [15,31]. miR-15a and miR-16 downregulation contributes to MM pathogenesis by promoting cells growth and neoangiogenesis in bone marrow [31]. A high level of the miR-17-92 cluster is, on the other hand, associated with poor prognosis in patients with MM, and it has been speculated that other factors may counteract the chromosome 13 deletion effect leading to the overexpression of these miRNAs in MM, as discussed below [32]. Furthermore, miR-22, encoded at the 17p locus, was significantly less expressed in those patients with the 17p deletion. [33]. Recently, miR-22 has been identified as a tumor-suppressing miRNA and its expression is decreased in a variety of human neoplasms [34][35][36][37]. Moreover, the downregulation of 14 miRNAs in the presence of loss of heterozygosity (LOH) has been detected, such as let-7b at 22q, miR-662 at 16p or miR-140-3p at 16q [17]. A probable consequence in MM can arise from the fact that let-7 miRNAs (a family of 9 mature let-7, encoded by 12 different genomic loci (reviewed in [38])) function as a tumor suppressor through regulation of key oncogenes, including MYC and RAS [39,40]. Low expression of let-7 family members is associated with poor prognosis in several cancer types [41,42]. In human myeloma cell lines (HMCLs), 61% of miRNA gene loci were affected by CNAs, showing a predominance of gains versus losses [33].

microRNA Expression and Epigenetics
Epigenetic modifications have a strong impact on gene expression regulation. Genes with methylated CpG in the promoters are typically repressed, while genes with unmethylated promoters can be expressed or not depending on other mechanism of expression regulation (reviewed in [43]). In many tumors, changes of DNA methylation have been found to be correlated with disease stage and patient survival (reviewed in [44]). In MM, the global DNA methylation level was observed to decrease from monoclonal gammopathy of undetermined significance (MGUS) to MM, although without further changes between newly diagnosed and relapsed cases [45,46]. Moreover, methylation

microRNA Expression and Epigenetics
Epigenetic modifications have a strong impact on gene expression regulation. Genes with methylated CpG in the promoters are typically repressed, while genes with unmethylated promoters can be expressed or not depending on other mechanism of expression regulation (reviewed in [43]). In many tumors, changes of DNA methylation have been found to be correlated with disease stage and patient survival (reviewed in [44]). In MM, the global DNA methylation level was observed to decrease from monoclonal gammopathy of undetermined significance (MGUS) to MM, although without further changes between newly diagnosed and relapsed cases [45,46]. Moreover, methylation of DNA encoding particular miRNAs has also been correlated with cancer progression. Increased methylation of the promoter regions of miR-34a [47], miR-152, miR-10b-5p [48] and miR-203 was observed in MM cells compared to normal plasma cells [49]. In this regard, hypermethylation of miR-34b/c and miR-129-2 has been detected during the transformation from MGUS to MM, as well as in MM relapse/progression [50,51]. Moreover, miR-375 and miR-342-3p hypermethylation was also found in relapsed/progressed MM [52,53].
Conversely, some miRNAs can directly affect epigenetic machinery by modulating the expression of enzymes, such as DNA methyltransferases (DNMT1, DNMT3a and DNMT3b) or histone methyltransferases. These miRNAs are often considered as a family of epigenetic miRNAs, "epi-miRNAs" [54]. In MM, DNMT1 has been shown to be targeted by miR-152, whereas miR-126 and miR-140-3p decrease DNMT1 expression in other diseases [55,56]. Hypomethylation treatment or miR-152 transfection of MM cells resulted in the decreased expression of DNMT1, E2F3 [48]. The role of miR-126 and miR-140-3p has been investigated in the interplay between myeloma cells and microenvironment. Co-culturing WL2 MM cells with bone marrow stromal cells (BMSC) revealed increased expression of miR-126 and miR-140-3p, which led to repression of DNMT1 and upregulation of RANKL [57]. The effect of the microenvironment on the enzymes responsible for DNA methylation is further supported by the experiments demonstrating increased expression of two DNA methyltransferases, DNMT3A and DNMT3B, after co-culture of MM cells with BMSC [58]. Moreover, expression of DNMT3A and DNMT3B is downregulated by miR-29b [59]. Overexpression of miR-29b reduced general DNA methylation two-fold. Interestingly, expression of miR-29b was increased by downregulation of histone methyltransferase EZH2 [60]. miR-29b also targets the anti-apoptotic MCL-1 protein [61], CDK6, PSME4 [62] and HDAC4 [21], which further supports links among epigenetics modification, drug resistance and apoptosis. A summary of epigenetically regulated miRNA is presented in Table 1.

Transcriptional Regulation of miRNA Expression
Intragenic miRNAs, located within introns or exons of protein-coding genes, are co-transcribed by Pol II and their expression is generally regulated by the same mechanisms as those of the host genes [6,70]. Little evidence exists about the relationship between expression of miRNAs and their host genes in primary MM samples, but so far no significant correlation between the intronic miRNAs and their host transcript expression has been detected [15]. Moreover, analysis of microarray-based datasets in various cell types, including MM, revealed that the large majority of intragenic miRNAs are not co-expressed with their host genes, and in MM only 11% of them were significantly correlated [71]. In HMCLs however, several associations have been described and are summarized in Table 2. Table 2. miRNA co-regulated with their host genes and the mechanism described in MM. 17q23.1 miR-21 TMEM49

Gene
By an upstream enhancer, overlapping with intronic region of the host gene [73] 21q21.3 miR-155 BIC By promoter DNA methylation of the host gene [68] The intergenic miRNAs are transcribed from their own promoter by Pol II or Pol III, and a small group of miRNAs can be transcribed without any exogenous promoter [6,70,74]. Several transcription factors (TFs) have been described to regulate the intergenic miRNAs expression in MM ( Figure 2).
A number of studies indicate that the bone marrow microenvironment (BMM) induces growth-promoting effects by modulating the expression of miRNAs [75,76]. The first published work regarding the regulation of miRNA expression in MM showed that interleukin 6 (IL6), the cytokine produced by bone marrow mesenchymal stem cells, induced changes in miRNA expression enhancing survival and drug resistance of myeloma cells. One of the miRNAs induced by IL6 is miR-21, whose stimulation is mediated by an upstream enhancer, the regulatory site that contains STAT3 binding sites. STAT3-mediated induction of miR-21 has a pro-survival effect on MM cells, since ectopic increase of miR-21 in IL6-dependent MM cell lines abolished the dependency on this cytokine [73]. STAT3 has been previously shown to induce the expression of anti-apoptotic genes, like survivin, Bcl-2 or MCL1 [77,78]. Moreover, downregulation of IL6 by treatment with berberine (a natural alkaloid derived from a traditional Chinese herbal medicine) decreased miR-21 transcription through STAT3 down-regulation, which further validate the prosurvival impact of miR-21 in MM [73]. Additional evidence of miR-21 regulation by STAT3 was provided using ibrutinib (BTK inhibitor), which inhibits miR-21 transcription by disrupting NF-kB and STAT3 binding to the miR-21 promoter [79]. Moreover, BMM is highly hypoxic, which leads in MM cells to an AKT-mediated decrease of miR-199-5p expression, and consequently to increased expression of HIF-1α [80], a transcription factor involved in hypoxia response [81]. Table 2. miRNA co-regulated with their host genes and the mechanism described in MM. By an upstream enhancer, overlapping with intronic region of the host gene [73] 21q21.3 miR-155 BIC By promoter DNA methylation of the host gene [68] The intergenic miRNAs are transcribed from their own promoter by Pol II or Pol III, and a small group of miRNAs can be transcribed without any exogenous promoter [6,70,74]. Several transcription factors (TFs) have been described to regulate the intergenic miRNAs expression in MM ( Figure 2).  C-Myc is a transcription factor with recognized oncogenic functions. In MM, c-Myc deregulation is one of the key events associated with disease progression (reviewed in [82]). In order to regulate the transcription of the genes, c-Myc binds to DNA at multiple regulatory sites, one being the so-called E-Box. This evolutionary conserved DNA site was detected in the host gene of miR-155 [68], one of the miRNAs essential for the B-lineage lymphocyte [83,84]. In fact, it has been demonstrated that c-Myc, Mad1 and Max regulate the expression of miR-155 in MM by binding to this E-Box site [68]. In addition to miR-155, c-Myc activation results in widespread reprogramming of the miRNA expression pattern of tumor cells [85]. miRNAs belonging to the miR-17-92 cluster were found to be significantly up-regulated only in MM samples but not in MGUS or healthy PCs [16] and the expression of this cluster is regulated by c-Myc in MM [86]. In this sense, it has been described that c-Myc inhibits apoptosis of MM cells by activation of miR-17-92 cluster, which leads to the downregulation of the proapoptotic protein Bim [86]. Other targets of that miRNA cluster are PTEN and E2F1, which support its tumor-promoting functions [87][88][89][90][91]. Recently, the miR-17-92 cluster was targeted with gapmeR (a chimeric antisense oligonucleotide that contains a central block of deoxynucleotide monomers sufficiently long to induce RNase H cleavage) antisense oligonucleotides, MIR17PTi, to induce degradation of MIR17HG primary transcripts and consequently prevent biogenesis of these miRNAs. As a result, MIR17PTi induced cell growth impairment, apoptosis, senescence and sensitized MM cells resistant to dexamethasone, melphalan or bortezomib. Moreover, treatment with MIR17PTi induced death only in cells with active MYC, through the upregulation of proapoptotic protein BIM and tumor suppressors p21, TP53, E2F1 and PTEN [92]. This finding demonstrates the myc-dependent synthetic lethality of this compound in MM.
On the other hand, c-Myc takes part in a multi-component regulatory complex which trans-repress several miRNAs in MM, including miR-23b, miR-29b or miR-29a [75,93,94]. The repression of miR-23b and miR-29b in MM leads to activation of Sp1, a transcription factor that regulates the expression of cell-cycle, differentiation, and apoptosis-related genes [93]. Many miRNAs and transcription factors form autoregulatory loops, in which they mutually regulate each other [95]. A miR-29b-Sp1 feedback regulatory loop was described in MM [96]. Moreover, the treatment of MM cells with bortezomib decreased the expression of Sp1 with the consequent increase of miR-29b level, indicating that this miRNA can be regulated not only by c-Myc but also by Sp1 [75].
TP53 is a tumor suppressor that interacts with numerous pathways and is commonly mutated in various cancers. However, in MM TP53 mutations are not so frequent and p53 seems rather to be inactivated by other multiple factors (reviewed in [97]). Recent studies have demonstrated that miRNAs interact with p53 and this network regulates the p53 level, and, in turn, p53 also regulates the transcription, expression and the maturation of a group of miRNAs [28,98,99]. In MM, an autoregulatory loop between the miR-194-2-192 cluster and p53 has been discovered, in which p53 acts as a transcriptional activator of pri-miR-194-2 by directly binding to its core promoter element, and those miRNAs directly target 3 UTR of TP53 [28]. Moreover, BMM induces expression of miR-125-5p in MM, leading to decrease of p53 expression and consequently, decrease of the p53-regulated miR-194-2-192 cluster [100].
The knowledge of transcriptional regulation of miRNAs expression will help us better understand not only the function of miRNAs but also the effect of deregulation of transcription factors in MM.

Impaired miRNA Processing may Deregulate miRNA Expression
The biologically active miRNAs are generated in a two-step sequential mechanism involving two RNase III nucleases, Drosha and Dicer. The downregulations of these key factors in the biogenesis of miRNA has been reported in many cancer types (reviewed in [101]). Processing of the hairpin precursor (pre-miRNA) through Dicer generates a miRNA duplex, consisting of a miRNA (5p strand) and miRNA* (3p strand). The miRNA machinery is orchestrated by two major multiprotein complexes, the Drosha complex in the nucleus, and the cytoplasmatic RNA-induced silencing complex (RISC), which contains the Argonaute (Ago) family proteins as a core component (Figure 3) (reviewed in [12,14,[102][103][104][105][106][107]). One of these strands generated by Dicer, known as the guide arm, will be incorporated into the RISC complex while the other, known as the passenger strand, will be degraded. One of the two strands is preferentially selected as a guide, either 5p or 3p. While either of the miRNAs strands can be used equally, the selection of the strand is highly regulated and depends on cell/tissue type [71,108]. The relationship between expression of sister miRNAs was compared in MM and other pathologies, such as acute lymphoblastic leukemia and prostate cancer, and the ratio between expression values of sister miRNAs was able to distinguish MM from other tissues [71]. The two mature miRNAs (3p and 5p) have partially reversed complementary sequences and tend to have different targeting properties [109].
Non-Coding RNA 2019, 5, x 7 of 20 autoregulatory loop between the miR-194-2-192 cluster and p53 has been discovered, in which p53 acts as a transcriptional activator of pri-miR-194-2 by directly binding to its core promoter element, and those miRNAs directly target 3′UTR of TP53 [28]. Moreover, BMM induces expression of miR-125-5p in MM, leading to decrease of p53 expression and consequently, decrease of the p53-regulated miR-194-2-192 cluster [100]. The knowledge of transcriptional regulation of miRNAs expression will help us better understand not only the function of miRNAs but also the effect of deregulation of transcription factors in MM.

Impaired miRNA Processing may Deregulate miRNA Expression
The biologically active miRNAs are generated in a two-step sequential mechanism involving two RNase III nucleases, Drosha and Dicer. The downregulations of these key factors in the biogenesis of miRNA has been reported in many cancer types (reviewed in [101]). Processing of the hairpin precursor (pre-miRNA) through Dicer generates a miRNA duplex, consisting of a miRNA (5p strand) and miRNA* (3p strand). The miRNA machinery is orchestrated by two major multiprotein complexes, the Drosha complex in the nucleus, and the cytoplasmatic RNA-induced silencing complex (RISC), which contains the Argonaute (Ago) family proteins as a core component ( Figure 3) (reviewed in [12,14,[102][103][104][105][106][107]). One of these strands generated by Dicer, known as the guide arm, will be incorporated into the RISC complex while the other, known as the passenger strand, will be degraded. One of the two strands is preferentially selected as a guide, either 5p or 3p. While either of the miRNAs strands can be used equally, the selection of the strand is highly regulated and depends on cell/tissue type [71,108]. The relationship between expression of sister miRNAs was compared in MM and other pathologies, such as acute lymphoblastic leukemia and prostate cancer, and the ratio between expression values of sister miRNAs was able to distinguish MM from other tissues [71]. The two mature miRNAs (3p and 5p) have partially reversed complementary sequences and tend to have different targeting properties [109]. The application of deep sequencing technologies to short RNAs and the analyses of pre-miRNA processing have led to the discovery of a distinct class of small RNAs known as moRNAs (miRNAoffset RNAs) derived from the sequence flaking 5'-or 3´arm pre-miRNA [110]. MoRNAs found in human tissues are generally included in the miRNA hairpin precursor [111]. Although the biogenesis The application of deep sequencing technologies to short RNAs and the analyses of pre-miRNA processing have led to the discovery of a distinct class of small RNAs known as moRNAs (miRNA-offset RNAs) derived from the sequence flaking 5 -or 3 arm pre-miRNA [110]. MoRNAs found in human tissues are generally included in the miRNA hairpin precursor [111]. Although the biogenesis and function of moRNAs remains unclear, it has been hypothesized that moRNA biogenesis might be associated with pre-miRNA processing [112,113]. Some data indicate that moRNAs may repress target transcripts as miRNAs do [114]. On the other hand, nuclear moRNA enrichment found in metazoan cells might indicate their role in regulating epigenomic modifications and transcription [115]. Likewise, the function of moRNAs in MM remains to be elucidated.
The ubiquitously expressed ubiquitin E3 ligase protein, cereblon, has been identified as a direct target of antimyeloma immunomodulatory drugs, like lenalidomide [116,117]. Remarkably, the Ago2 protein was found to be a substrate for the cereblon protein. Upon treatment of MM cells with lenalidomide, the level of Ago2 protein decreased with the consequent downregulation of global miRNA expression level. In line with this, the knocking down of AGO2 and Dicer significantly decreased growth and viability of myeloma cells [118]. Moreover, silencing of AGO2 led to decrease of the total miRNA level [118,119]. Moreover, Ago2 is involved in promoting angiogenesis by miRNA-mediated regulation of pro-and anti-angiogenic signals [120,121]. Expression of AGO2 mRNA, however, did not significantly associate with global miRNA expression in primary MM cells [118]. Dicer mRNA level in NPC was very similar to MGUS, a pre-malignant condition, and significantly higher than in smoldering myeloma (SMM) and MM [122]. Moreover, higher expression of Dicer was associated with improved progression-free survival in symptomatic myeloma patients [122].
Dicer silencing led to cell cycle inhibition in MM cell lines [118], while higher Dicer expression was not associated with cell-cycle activation of proliferation in primary MM cells [122]. These results suggest that the effect of Dicer downregulation might be different in cell lines compared to the pathophysiological effect in primary MM. It is noteworthy that the miRNA expression level in MM is also affected by mutual crosstalk between myeloma cells and BMM, leading to the deregulation of certain miRNAs in MM [76].
Overall, these studies support that idea that the altered expression of proteins involved in the biogenesis of miRNAs may affect the MM pathogenesis.

Impact of miRNA-mRNA Associations
It is estimated that miRNAs repress the translation of approximately 30-50% of all the coding genes [123]. The effect of miRNAs usually induces a negative correlation between their expression levels and their target genes. However, this is not always the case and sometimes the regulatory effect is only observed at translation, with a low effect on the mRNA transcript levels [124,125]. miRNA-mediated gene regulation has a broad impact on gene expression since a single miRNA can control many transcripts and, in turn, a single transcript can be regulated by many miRNAs [126]. All these factors mean that the accurate identification of the targets regulated by miRNAs remains a challenge. Traditionally, miRNA-mRNA associations have been predicted using bioinformatic algorithms based on the sequence conservation between the miRNA and the target gene, such as miRanda [127], PicTar [128] or TargetScan [129]. More sophisticated approaches include other parameters such as the binding free energy and the secondary structure of the 3 UTR of the target gene. These improvements were implemented in tools such as PITA [130] or rna22 [131]. Nevertheless, even though these methods can identify the regulatory potential of miRNAs, it is noteworthy that they cannot reach the miRNA-targeted mRNA pairs with biological significance, as it depends on the association between the expression levels of both molecules and the biological context [132]. In recent years, the use of high-throughput techniques, such as microarrays, has enabled the detection of potential associations between miRNA and mRNA levels. However, these approaches alone are still limited at the time of revealing the mechanisms involved in the transcriptional cascade and how they are regulated. Many bioinformatic tools have been developed to integrate all these data through the generation of co-expression networks, using methods based on correlation [133] or Bayesian inference [134,135] or "mixed regulatory circuits reconstruction" [136]. These networks are built from expression data at both miRNA and mRNA level in order to identify critical genes in the network structure and to determine new regulatory relationships. Genes and miRNA are represented as the nodes of the network and the edge that connects two nodes represents the relationship between them. The aim of these methods is to detect the critical elements in the network, identifying the essential nodes and edges in the network structure. In summary, methods based on miRNA and mRNA expression integration have the major advantage of considering the biological background of the studied samples, however, all the strategies mentioned above require further validation. The in silico approach for validation is to use and compare different algorithms to predict the miRNA-mRNA interactions [137,138], which is widely recommended before the target validation experiments [138]. Several experimental strategies have been performed to determine the truthfulness and the biological significance of the miRNA-targeted mRNA interaction [139][140][141]. The most commonly used strategies are the validation of the expression of the miRNA and its targeted mRNA through qRT-PCR or Northern blot, and the demonstration of the miRNA interaction with a target site using the luciferase reporter assay [141]. However, although the use of these approaches is much extended in the literature, both have their pros and cons. The main disadvantage of co-expression assays is that they cannot differentiate direct and indirect miRNA targets [140], so it is not possible to know whether a miRNA directly interacts with an mRNA, or whether this effect is mediated by another target molecule. In the case of the luciferase reporter assay, the main drawbacks are that they depend on the cloning region and the variability of the protocol [140]. Because of this, it is necessary to combine different validation strategies to effectively determine the miRNA-mRNA interactions [140].
The expression of the miRNA and its targeted mRNA has also been actively studied in MM [15,17,132,142,143]. In silico analysis of the relationship between the miRNA and mRNA expression profiling in the context of the different MM cytogenetic subgroups has shown interactions involving genes related to biological pathways that could play a relevant role in the development of MM [15]. In the t(4;14) subgroup, miR-135b and miR-146a targeted PELI2 and IRAK1 genes, which are involved in the IL1 signaling pathway. A batch of four genes (GPD2, GLCCI1, FNDC3B and ASH2L) previously reported in cancer development were found to be targeted by the miR-1 in the t(14; 16) subgroup. Concerning the 13q deletion, the potential regulation of the MAPK pathway by miRNAs was highlighted. A final comparison among all cytogenetic subgroups revealed that some miRNA-mRNA interactions involving the CCND2 gene were commonly dysregulated in the t(4;14), t(14;16) and 13q deletion subgroups, suggesting the potential targeting of CCND2 by miRNAs in MM. Interactions were considered true positive when they were predicted by at least three out of the four algorithms used for the integrative analysis: TargetScan, miRDB, miRanda and Pictar. A similar study was performed through an integrative analysis with a correlation network of functional interactions [17]. This study identified more than 23,000 regulatory relationships involving 628 miRNAs and 6435 genes, which were considered real interactions if they were within the top 3% of the anti-correlated miRNA-targeted mRNA pairs. The most consistent sub-network was the corresponding to the t(4;14), which grouped seven miRNAs and 289 targets. In this sub-network, three of the target genes (CBFA2T2, PP1R16B and GOSR2) were commonly regulated by five miRNAs.

Single Nucleotide Polymorphisms and 3 UTR Polyadenylation Affect miRNA Binding Sites
The function of miRNAs may also be affected by sequence variations in miRNA binding sites, such as mutations and single nucleotide polymorphisms (SNPs) in the 3 UTR. These genetic variations inside 3 UTRs may overlap with miRNA binding sites and impair the translation inhibition or degradation of the mRNAs, or create new miRNA binding sites [144]. Computational analysis has identified more than 400 SNPs located within predicted and experimentally verified miRNA-binding sites [25]. Moreover, in recent years several studies have shown associations between SNPs in miRNA-binding-sites (miRSNPs) and cancer (reviewed in [145]). The first evidence for miRSNPs associated with hematological malignancies arise from a study in MM, in which the prognostic impact of six miRSNPs, located either in miRNAs target genes or in miRNA biogenesis pathway proteins, was evaluated. Only two miRSNPs, rs3660 in the KRT81 gene and rs11077 in the XPO5 gene, which were experimentally validated using renilla/luciferase reporter assays, were associated with better prognosis in MM after autologous stem cell transplantation [146] (Figure 4, left side).
On the other hand, another study based on computational prediction using genome-wide analysis and meta-analysis found five SNPs associated with MM risk [147] (Figure 4, right side). Cancer cells often express substantial amounts of mRNA isoforms with shorter 3'UTR regions that usually result from alternative cleavage and polyadenylation (APA) [148]. The loss of miRNAmediated repression due to the alternative cleavage and APA enhances the generation of shorter mRNAs isoforms that exhibit higher stability and produce more proteins [148,149]. This mechanism seems to be involved in the overexpression of CCND2 in multiple myeloma. In fact, the shortening of CCND2 3′UTR by alternative polyadenylation with the consequent loss of miRNA binding sites was demonstrated both in myeloma cell lines and primary myeloma samples [150].

Regulation of miRNA Expression by miRNA Sponges
miRNA activity can be affected not only by the aforementioned factors, but also by regulatory RNA species, the so-called competing endogenous RNA (ceRNAs) or miRNA 'sponges'. Sponge RNAs contain binding sites for miRNAs, and hence compete with target mRNAs for miRNA binding ( Figure 5). RNAs that may function as sponges include long noncoding RNAs (lncRNAs), RNAs encoded by pseudogenes and circular RNAs (reviewed in [151]). Furthermore, highly abundant mRNAs have also been described to serve as endogenous sponges, since they may sequester miRNAs and suppress their activity over other mRNA targets. In general, miRNAs are stable and destroy targeted mRNAs. However, research suggests that endogenous targets may also control miRNAs by target RNA-directed miRNA degradation (TDMD) [152]. Interestingly, it has been described that some miRNAs can induce the degradation of specific circular RNAs after binding, whereas others are resistant and therefore act as strong miRNA sponges [153]. The ability to act as a miRNA sponge is also increased by the presence of multiple tandem high-affinity binding sites to a microRNA. Thus, the circular RNA sponge for miR-7 (ciRS-7) contains >60 conserved miR-7 seed matches and, therefore, can bind densely to this miRNA [154]. Cancer cells often express substantial amounts of mRNA isoforms with shorter 3 UTR regions that usually result from alternative cleavage and polyadenylation (APA) [148]. The loss of miRNA-mediated repression due to the alternative cleavage and APA enhances the generation of shorter mRNAs isoforms that exhibit higher stability and produce more proteins [148,149]. This mechanism seems to be involved in the overexpression of CCND2 in multiple myeloma. In fact, the shortening of CCND2 3 UTR by alternative polyadenylation with the consequent loss of miRNA binding sites was demonstrated both in myeloma cell lines and primary myeloma samples [150].

Regulation of miRNA Expression by miRNA Sponges
miRNA activity can be affected not only by the aforementioned factors, but also by regulatory RNA species, the so-called competing endogenous RNA (ceRNAs) or miRNA 'sponges'. Sponge RNAs contain binding sites for miRNAs, and hence compete with target mRNAs for miRNA binding ( Figure 5). RNAs that may function as sponges include long noncoding RNAs (lncRNAs), RNAs encoded by pseudogenes and circular RNAs (reviewed in [151]). Furthermore, highly abundant mRNAs have also been described to serve as endogenous sponges, since they may sequester miRNAs and suppress their activity over other mRNA targets. In general, miRNAs are stable and destroy targeted mRNAs. However, research suggests that endogenous targets may also control miRNAs by target RNA-directed miRNA degradation (TDMD) [152]. Interestingly, it has been described that some miRNAs can induce the degradation of specific circular RNAs after binding, whereas others are resistant and therefore act as strong miRNA sponges [153]. The ability to act as a miRNA sponge is also increased by the presence of multiple tandem high-affinity binding sites to a microRNA. Thus, the circular RNA sponge for miR-7 (ciRS-7) contains >60 conserved miR-7 seed matches and, therefore, can bind densely to this miRNA [154].
target RNA-directed miRNA degradation (TDMD) [152]. Interestingly, it has been described that some miRNAs can induce the degradation of specific circular RNAs after binding, whereas others are resistant and therefore act as strong miRNA sponges [153]. The ability to act as a miRNA sponge is also increased by the presence of multiple tandem high-affinity binding sites to a microRNA. Thus, the circular RNA sponge for miR-7 (ciRS-7) contains >60 conserved miR-7 seed matches and, therefore, can bind densely to this miRNA [154].  The influence of miRNA sponges in the regulation of cancer-associated genes is nicely illustrated by PTEN, one of the most frequently disrupted tumor suppressors in cancer [155]. Several protein-coding ceRNAs have been identified that share miRNA target sites with PTEN. Moreover, PTEN levels are also regulated by a PTEN pseudogene, PTENP1, which acts as a ceRNA by providing additional miRNA target sites [156]. Deregulation of these PTEN-ceRNAs seems to be responsible for PTEN under-expression in some tumors [157]. Other ceRNAs that influence cancer-related factors have been described [156], such as LncRNA CRNDE, which is overexpressed in colorectal cancer and promotes cell proliferation and metastasis [158,159]. Overexpression of CRNDE has also been found in MM and has been related to tumor progression and poor survival [160]. The authors reported that CRNDE induced the proliferation and an antiapoptotic response in MM by sponging miR-451. The lncRNAs, CCAT1 and MALAT1, which are also overexpressed in MM [161][162][163], have been shown to promote MM cell growth by functioning as ceRNA for miR-181a-5p and miR-509-5p, respectively [161,164]. The lowered availability of these miRNAs increased the abundance of the tumor-promoting factors HOXA1 and FOXP1, respectively.
In analogy with naturally occurring miRNA suppression mechanisms, artificially custom-designed sponges can be exploited for inhibition of miRNA activity (Reviewed in [165]). In fact, knockdown of overexpressed oncogenic miRNAs with this technology can be a therapeutic strategy for cancer [166]. Thus, in MM it has been reported that inhibition of miR-19 and miR-155 activity with "traditional" or artificially-improved sponges increases SOCS1 levels, leading to enhanced activity of tumor suppressor p53 and inhibition of the transformed phenotype [167]. The deepest knowledge of miRNA sponges will pave the way to manipulate miRNA functions in MM and other cancer cells.

Conclusions
A full understanding of the role of miRNAs in MM pathogenesis requires not only exploration of the different effects of the deregulation of mature miRNAs, but also unveiling of the factors that regulate their expression and ultimately determine their function. These factors encompass the genomic modifications of DNA sequences encoding miRNAs, including DNA methylation, which is one of the most significant mechanisms that regulate miRNA expression in MM, the transcriptional regulation of miRNAs and, finally, variations in the mRNA target sites. Moreover, other elements, such as competing endogenous RNA and alterations in the complex enzymatic machinery that generate the mature and active miRNAs, may affect their final expression.
The understanding of the complex mechanisms of regulation that determine the expression and functionality of the mature miRNAs in MM may open new avenues for investigation, particularly in the field of therapeutic interventions.