DNA Methylation of Tumor Suppressive miRNAs in Non-Hodgkin’s Lymphomas

DNA methylation is an epigenetic alteration leading to heritable phenotypic changes of cells with functional consequences. It is important in early embryonic development, stem cell differentiation, and tissue-specific gene expression. In normal cells, promoter-associated CpG islands (CGI) are generally unmethylated except in X-chromosome inactivation or genomic imprinting. In cancer, tumor cells are characterized by global hypomethylation but locus-specific hypermethylation of promoter-associated CGI, resulting in gene silencing. MicroRNAs (miRNAs) are short, non-coding RNA sequences of 18–25 nucleotides, which can repress the translational of multiple protein-coding mRNAs by sequence-specific binding to the 3′untranslated region. Depending on the genes targeted, miRNA can be tumor suppressive if an oncogene is repressed, or it can be oncogenic when a tumor suppressive gene is repressed. Recently, aberrant methylation of tumor suppressive miRNAs has been reported in different types of cancers including lymphomas. Herein, we review the recent literature of methylation of tumor suppressive miRNAs in different histopathologic subtypes of lymphomas, and discuss its potential diagnostic, prognostic, and therapeutic significance.


LYMPHOMA
Lymphoma results from neoplastic proliferation of lymphocytes. According to the World Health Organization classification, it can be broadly classified into Hodgkin and non-Hodgkin lymphomas (NHL; Swerdlow et al., 2008). Hodgkin lymphoma is characterized by CD30+ve Reed Sternberg cells with variable degree of reactive cellular infiltrate or sclerosis. NHLs are more heterogeneous, and are generally classified by the cell lineage, histological pattern, morphology, immunophenotype, or maturity of lymphoma cells. Based on the lineage of lymphoma cells, NHLs can be classified as B-, T-, or natural killer (NK)-cell lymphomas. For Band T-cell lymphoma, they are further classified as precursor Bor T-cell lymphoid neoplasms, or mature B-or T-cell neoplasms (Chan, 2001;Au et al., 2005; Table 1). NHLs often present as nodal enlargements. However, extranodal sites such as the bone marrow, liver, spleen, mediastinum (Chim et al., 1996), heart (Chim et al., 1997), central nervous system, testis, breast, and the gastrointestinal tract can be involved. Moreover, some NHLs present primarily as extranodal disease. A prominent example is marginal zone Bcell lymphoma (MZBCL) of the stomach, which is an extranodal MZBCL of mucosa-associated lymphoid tissue (MALT) frequently associated with Helicobacter pylori infection of the gastric mucosa (Chan, 2001). Interestingly, in H. pylori+ve MALT lymphoma of the stomach, eradication of H. pylori infection alone may lead to resolution of gastric lymphoma.
Mature B-cell lymphomas can often be conceptually grouped by the putative maturation ontogeny of the neoplastic cells. Lymphomas arising from transformation of germinal center Bcells, which are CD10+ve, include follicular lymphoma (FL), Burkitt's lymphoma (BL), and some diffuse large B-cell lymphoma (DLBCL). DLBCL is the most common form of mature B-cell lymphoma, comprising about 30% of all NHL. Despite its clinical aggressiveness, with combination chemotherapy, about half of the patients may be cured. On the other hand, small B-cell lymphomas comprise FL, small lymphocytic lymphoma (SLL), MZBCL, BL, lymphoplasmacytic lymphoma, and mantle cell lymphoma (MCL). FL is one form of small B-cell lymphoma prevalent in the Western population, accounting for about 20% of NHL. FL is characterized by the presence of t (14;18) with upregulation of BCL2, thereby conferring survival benefit to the lymphoma cells. It is an indolent disease of the elderly, and often presents with nodal and advanced stage disease. SLL is the nodal counterpart of chronic lymphocytic leukemia (CLL), and the tumor B-cells express dually CD5 and CD23 in addition to pan-B-cell antigens, CD19, and CD20. MZBCL, another form of indolent, small B-cell lymphoma, may occur as a nodal or splenic lymphoma in addition to the extranodal presentation described above.
On the other hand, despite being a form of small B-cell lymphoma, BL is an extremely aggressive lymphoma, characterized histologically by a starry sky appearance with literally all tumor cells simultaneously engaged in cell proliferation as evidenced by the almost 100% Ki67 immunoreactivity, and cytogenetically by t (8;14) in the majority of cases, which leads to upregulation of the MYC oncogene at 8q24. BL was first discovered in Africa as an extranodal lymphoma with a high proliferation rate associated with Epstein-Barr virus (EBV) infection. Subsequently, sporadic BL, often with extranodal presentation, had been diagnosed in other parts of the world.  Mantle cell lymphoma, once thought to be an indolent small B-cell lymphoma, is clinically moderately aggressive. It is characterized by t (11;14), leading to the uniform upregulation of cyclin D1 (Chim et al., 1998). MCL often presents with advanced disease with involvement of the lymph nodes and frequent extranodal involvement including spleen, blood, and bone marrow. (Chim et al., 2003a) With standard therapy, MCL is associated with a short remission duration with a median overall survival of 4-5 years (Vose, 2012).
It is notable that T-cell lymphoma is much less frequent than B-cell lymphoma, and depending on the geographic location, it comprises only 8-15% of all NHLs (Au et al., 2005). Anaplastic large cell lymphoma (ALCL), peripheral T-cell lymphoma, unspecified, and angioimmunoblastic T-cell lymphoma (AITL) are the most common forms of mature PTCL (Au et al., 2005). Most forms of mature T-cell lymphomas are nodal lymphomas. Of these, the pathology of PTCL, unspecified, were variably described as T-zone lymphoma, lymphoepithelioid cell lymphoma, pleomorphic T-cell lymphoma, small, medium, or large sized types, and T-immunoblastic lymphoma; so that there was a broad morphologic spectrum and immunophenotypic profiles. On the other hand, AITL is a disease of the elderly, with the majority being nodal and advanced stage at presentation. Pathologically, there was a polymorphic lymphomatous infiltrate admixed with reactive lymphocytes, eosinophils, plasma cells, and histiocytes, with arborizing high endothelial venules. Systemic manifestations includes hepatosplenomegaly, ascites, proteinuria, and high fever (Chan, 2001;Au et al., 2005). Finally, in ALCL, lymphoma cells are large and pleomorphic, expressing CD30 in addition to T-cell markers. ALCL can further be classified by the expression of the anaplastic lymphoma kinase (ALK) into ALK-positive and ALK-negative subtupes. In ALK-positive ALCL, ALK is often activated as a result of reciprocal translocation of the ALK gene with another partner gene, with t (2;5) being the prototype that results in fusion of the nucleophosmin gene with ALK. ALK-positive ALCLs carry a favorable prognosis compared with ALK-negative ALCLs.
Extranodal NK/T-cell lymphoma, nasal type, is derived putatively from NK-cells which are often CD56-positive. The majority of NK-cell lymphomas present primarily in the nasal area and upper aerodigestive tract, and is referred to clinically as nasal NK-cell lymphomas (Chim et al., 1999(Chim et al., , 2004c. Occasionally, NKcell lymphomas can arise in extranasal sites, involving the skin, gastrointestinal tract, salivary glands, and testis, and are referred to as non-nasal NK-cell lymphomas. Rarely, NK-cell lymphomas can present in a leukemic phase, and are referred to as NK-cell leukemia (Kwong, 2005). NK-cell neoplasms are rare but aggressive lymphomas uniformly associated with EBV infection. Patients with nasal NK-cell lymphoma often present with fever and nasal symptoms, and locally destructive lesions. Despite the initial nasal presentation, lymphoma cells may eventually disseminate to other extranodal sites, which often are the presentation sites of nonnasal NK-cell lymphomas. Finally, NK-cell leukemia presents with pancytopenia, bone marrow failure, and is almost always lethal.

miRNA
With complete sequencing of the human genome, an increasing amount of non-coding RNAs (ncRNAs) are unveiled and shown to work in concert with the protein-coding gene network (Human Genome Sequencing, 2004;Lander, 2011). miRNA, which is a widely studied subclass of short ncRNAs with a length of 18-25 nucleotides, leads to translational repression of proteincoding genes via sequence-specific binding of its seed region to the 3 untranslated region (UTR) of its target protein-coding genes. Currently more than 1500 miRNAs have been discovered. Interestingly, many miRNAs have been implicated in carcinogenesis (Kozomara and Griffiths-Jones, 2011). Based on the genomic location, miRNA can be classified as intergenic and intragenic. Intergenic miRNA is transcribed from non-coding region in between protein-coding genes, whereas intronic miRNA is encoded in the intron of, and mostly transcribed in parallel with, its host protein-coding gene (Lopez-Serra and Esteller, 2012).

Frontiers in Genetics | Non-Coding RNA
Similar to the transcription of protein-coding genes, most miRNA genes are also transcribed by RNA polymerase II with the inclusion of 5 cap structure and 3 polyadenylated tail in each of the primary miRNA (pri-miRNA) transcript (Lee et al., 2004). Pri-miRNAs, ranging from 100 to 1000 nucleotides in length, are then processed by a ribonuclease-III DROSHA complex with DGCR8 (DiGeorge Syndrome Critical Region Gene-8) into a stem-looped precursor miRNA (pre-miRNA; Gregory et al., 2004). These intermediates pre-miRNAs are exported via Ran-GTP-dependent exportin-5 (XPO-5) into the cytoplasm, where these pre-miRNA stem-loops are further processed into mature miRNA duplex (Yi et al., 2003). Eventually a single-stranded mature miRNA is produced, ready to function when it is loaded onto the DICER1-TAR RNA-binding protein-containing RNA-induced silencing complex (RISC; Liu et al., 2004). The biosynthesis and processing of miRNA is summarized in Figure 1.
Several functional mechanisms of miRNA have been described. However, the specific binding of miRNA seed region to the 3 UTR of mRNA via imperfect complementarity (Lopez-Serra and Esteller, 2012) has attracted much attention. Such binding leads to blockage of protein translation and hence translational silencing of the target protein-coding gene. Depending on the silenced target gene, miRNA can also be divided into tumor suppressive and oncogenic miRNA. A tumor suppressive miRNA targets an oncogene while an oncogenic miRNA targets a tumor suppressive gene. In lymphoma or solid cancers, in which viral infections is involved in carcinogenesis, potentially oncogenic miRNA may be derived from the integrated viral genome (Lin et al., 2010;Brown et al., 2012). Based on miRNA array studies, a multitude of tumor suppressive miRNAs has been shown to be downregulated in different types of lymphomas (Craig et al., 2011;Dejean et al., 2011;Iqbal et al., 2012). The primary objective of this article is to review the intricate role of tumor suppressive miRNAs that are silenced in lymphoma, and to discuss the emerging role of DNA methylation in mediating the expression of these miRNAs.

DNA METHYLATION-MEDIATED GENE SILENCING
DNA methylation is an important epigenetic mechanism for cells to maintain their normal gene expression patterns. Gene methylation is crucial for genomic imprinting and X-chromosome inactivation for correct tissue and organ development (Reik and Walter, 2001;Kaneda et al., 2004). DNA methylation occurring on CpG dinucleotides is mediated by DNA methyltransferases (DNMT)1, 3A, and 3B, which add a methyl group to the C5 carbon of the cytosine residue in the CpG dinucleotides (Okano et al., 1998). CpG rich regions, known as CpG islands (CGI), can be localized at the promoter region of a protein-coding or miRNA gene. CGI methylation results in a silencing effect on the associated protein-coding and miRNA genes. To determine the roles of CGIs in regulating gene expressions, genome-wide studies have been conducted to examine the methylation status of CpGs across the human genome in both normal and cancer cells. An account of methylation-mediated genomic imprinting and cancer-related transcriptional silencing is given below (Shen et al., 2007;Choi et al., 2010;Sandoval et al., 2011).

METHYLATION OF TUMOR SUPPRESSIVE miRNA IN NHL
All articles with the keywords "lymphoma, DNA methylation and miRNA" obtained from a search of the PubMed were reviewed. They can be classified into two groups. The first group comprised www.frontiersin.org mechanistic studies of miRNA methylation in special subtypes of NHL including NK-cell lymphoma (pertaining to miR-146a), gastric MALT lymphoma (miR-203), t (8;14)-negative BL (miR-9), and ALK-positive ALCL (miR-29a). The second group comprised studies of the impact of miRNA methylation in a panel of B-, T-, or NK-cell lymphomas (Table 2 and Figure 2).

miR-146a methylation in NK/T-cell lymphoma
Promoter methylation of miR-146a was described in a study of primary NK/T-cell lymphoma, which was associated with downregulation of miR-146a (Paik et al., 2011). Over-expression of miR-146a was then shown to inhibit lymphoma cell proliferation and induce apoptosis, thereby demonstrating its tumor suppressor properties. In addition, based on luciferase assay, overexpression of miR-146a was shown to lead to inhibition of the NFκB pathway, due to binding to and hence blockage of NFκB responsive elements, thereby illustrating a role of miR-146 on NFκB signaling. On the other hand, TNF receptor-associated factor 6 (TRAF6) has been shown to transactivate the NFκB pathway by promoting proteasomal degradation of IκB, the inhibitor of NFκB. By bioinformatic search, TRAF6 was found to possess  Frontiers in Genetics | Non-Coding RNA miR-146a binding sites at the 3 UTR, to which miR-146a binding has been demonstrated (Starczynowski et al., 2010). Besides, the authors showed that downregulation of TRAF6 was achieved by over-expression of miR-146a, consistent with the notion that TRAF6 is also a target of miR-146a (Paik et al., 2011). Moreover, downregulation of TRAF6 by siRNA led to inhibition of the NFκB pathway and consequent downregulation of anti-apoptotic BCL2, consistent with the notion that miR-146a regulated NFκB pathway via regulating TRAF6 expression. Furthermore, miR-146a was shown to confer chemosensitivity of NK lymphoma cells to etoposide, a chemotherapy active in NK/T-cell lymphoma. Finally, patients with higher miR-146a expression had better survival than those with low expression. Therefore, miR-146a is a putative tumor suppressive miRNA, acting via repression of TRAF6, and hence downregulation of NFκB signaling. Its frequent hypermethylation in NK/T-cell lymphoma might be of prognostic significance.

miR-203 methylation gastric MALT lymphoma
A recent study of miRNA expression in H. pylori-positive gastric MALT lymphoma showed differential expression of miRNA as compared with normal tonsils. In particular, miR-203 was shown to be downregulated in lymphoma (Craig et al., 2011), whereas the adjacent inflammatory but non-lymphoma tissues with chronic gastritis had higher expression levels. This was biologically relevant, as miR-203 expression was inversely related to the expression of the miR-203 target gene ABL1. Furthermore, miR-203 was unmethylated in normal tonsil, minimally methylated in gastritis tissue, but completely methylated in gastric MALT or DLBCL of stomach, which was derived from transformation of MALT lymphoma. Therefore, miR-203 methylation appeared specific to gastric lymphoma, which might have started with chronic gastritis. The tumor suppressive function of miR-203 was demonstrated in a mouse model, in which gastric MALT lymphoma developed after prolonged H. felis infection (Craig et al., 2011). In this model, miR-203 was highly expressed in normal marginal zone B-cells, modestly expressed in gastritis tissue and under-expressed in gastric MALT lymphoma tissue with corresponding upregulation of its target ABL1 proto-oncogene. Over-expression of miR-203 led to decreased ABL1 mRNA expression and a blockage of Helicobacter antigen-dependent cell proliferation. Finally, treatment with a tyrosine kinase inhibitor of imatinib that targeted ABL1 led to regression of gastric tumors (Craig et al., 2011). Therefore, miR-203 was shown to be a tumor suppressive miRNA targeting ABL1, and miR-203 methylation might be important in pathogenesis of MALT lymphoma.

miR-29a methylation in ALK+ve ALCL
A recent study showed that the expression of miR-29a was suppressed by promoter methylation in ALK-positive ALCL (Desjobert et al., 2011). Consistent with this observation, the myeloid cell leukemia 1 (MCL1) gene, a known target of miR-29a, was highly expressed in ALK-positive ALCL, whereas overexpression of miR-29a resulted in downregulation of MCL1 (Xiong et al., 2010;Desjobert et al., 2011). Interestingly, miR-29a expression could be restored either with knockdown of ALK or its inhibition, suggesting that ALK is involved in miR-29a methylation. Furthermore, knockdown of ALK or STAT3, known targets of miR-29a, led to downregulation of DNMT1 and DNMT3 (Desjobert et al., 2011). Therefore, ALK might repress miR-29a via upregulation of DNMT1 and hence hypermethylation of the miR-29a promoter. Finally, in a xenograft tumor model of ALK-positive ALCL, tumors with over-expression of miR-29a were much smaller that those without. These observations strongly suggested that miR-29a was a tumor suppressive miRNA hypermethylated in ALK-positive ALCL.

miR-9-1 methylation in t(8;14)-ve BL
The expression of mature miR-9 from the 3 arm of the three miR-9 precursors miR-9-1, miR-9-2, miR-9-3 (derived from different genomic loci) was low in t (8;14)-negative BL, as compared with DLBCL or t (8;14)-positive BL, thereby suggesting that miR-9-3p might be of pathogenetic significance in BL without t (8;14; Onnis et al., 2010). The low expression of miR-9-3p was found to be due to DNA hypermethylation of one of the three genomic loci, miR-9-1. The result was an upregulation of E2F1, a putative target of miR-9-3p. Since E2F1 is also a transcriptional target of c-MYC, miR-9-1 methylation with upregulation of E2F1 might be similar to the biological consequence of t (8;14), where c-MYC dysregulation might also result in excessive transcriptional activation of E2F1, leading in both instances to unchecked cellular proliferation (O'Donnell et al., 2005).

miR-203 methylation in NHL
Another study reported the status of miR-203 methylation in 150 patients with various hematological cancers (Chim et al., 2011b). In primary samples, miR-203 methylation was detected in 5.0% of ALL, 10.0% of AML, 42.0% of CLL, and 38.8% of NHL (including 60.0% of NK, 40.9% of B-cell, and 23.5% of T-cell NHL, and www.frontiersin.org hence miR-203 was more frequently hypermethylated in lymphoid than myeloid malignancies. Amongst the NHL samples studied, miR-203 was methylated in 40.9% of B-cell NHL, 23.5% of Tcell NHL, and 60% of NK/T-cell lymphoma. However, there was no correlation between miR-203 methylation with clinical parameters including age, gender, nodal/extranodal presentation, or Ann Arbor stage. Interestingly, in the patients in which other miR-NAs were concomitantly studied, miR-203 methylation was found to be associated with methylation of miR- 124-1, miR-34a, and miR-196b. As both miR-124 and miR-34a target CDK6, the simultaneous methylation of these miRNAs might contribute to constitutive activation of certain oncogenic pathways. Alternatively, inactivation of multiple tumor suppressive miRs targeting multiple pathways might be involved in lymphomagenesis. For instance, miR-203 methylation might contribute to tumor cell survival through upregulation of ABL1, and miR-124-1 methylation to proliferative advantage by upregulation of CDK6 (Martinez-Climent et al., 2001;Craig et al., 2011;Wong et al., 2011).

miR-34a methylation in NHL
Similarly, miR-34a methylation has been studied in a panel of hematological cancers including 20 ALL, 20 AML, 11 CML, 50 CLL, 55 MM, and 32 NHL patients. In primary samples at diagnosis, miR-34a methylation was detected in 4% of CLL, 5.5% of MM, and 18.8% of NHL at diagnosis, but not in ALL, AML, and CML (Chim et al., 2010). Amongst lymphoid malignancies, miR-34a was preferentially methylated in NHL, in particular NK/T-cell lymphoma. The findings suggested that mechanisms controlling the expression of a miRNA are cancer dependent, although miRNA might exert its tumor suppressive function in a similar mechanistic action in lymphoma of different subtypes.

DISCUSSION
Based on the above summary, a few points are worth further discussion.
Firstly, while mechanistic studies have been performed to illustrate the role of miR methylation in the pathogenesis of certain lymphoma subtypes (such as methylation of miR-146a in NK-cell lymphoma, miR-230 in gastric MALT lymphoma, and miR-29a in ALK+ve ALCL), it is likely that multiple miRs may be hypermethylated in any single lymphoma subtype. For instance, miR-203 have been shown to be hypermethylated in a multitude of hematological cancers especially CML and Ph+ve ALL (Bueno et al., 2008), and hence not restricted to gastric MALT lymphoma. This was further illustrated by our work that showed miR-203 methylation in multiple lymphoid malignancies including B-CLL, and B-cell, Tcell and even NK-cell lymphoma (Chim et al., 2011b). Therefore, multiple miRs are likely hypermethylated in a lymphoma subtype, and hence a lymphoma type-specific miR methylation profiles.
Thirdly, while some of these reported miRNAs, such as miR-9-1, 124-1, and -203, are embedded in classical CGI with proven promoter activities, some miRNAs were reported to be hypermethylated at scanty CpG sites upstream of the respective miRNA. For instance, miR-29a was postulated to be regulated by a stretch of 5 CpG sites at 1.1 kb upstream of its precursor sequence (Desjobert et al., 2011). Similarly, miR-146a was suggested to be regulated by two sparsely separated CpG dinucleotides immediately upstream of its precursor sequence, instead of the defined primary sequence (Paik et al., 2011). Last but not least, as DNMT has been shown to be the translational target of miR-29 family miRs, it would be interesting to see if restoration of miR-29a in hypermethylated cell lines may lead to re-expression of methylation-silenced tumor suppressor genes in addition to downregulation of DNMT family genes. For instance, miR-29b restoration in AML cells led to global DNA hypomethylation, resulting in re-expression of CDKN2B and estrogen receptor 1 (ESR1) through promoter DNA hypomethylation (Garzon et al., 2009). Moreover, in lung cancer, miR-29 family miRs (29a, 29b, and 29c) directly target both DNMT3A and -3B, and over-expression of miR-29s in lung cancer cell lines restores normal patterns of DNA methylation, induces re-expression of methylation-silenced tumor suppressor genes and inhibits tumorigenicity in vitro and in vivo (Fabbri et al., 2007).
In conclusion, there is much to be learned of miR methylation in NHL regarding the biology, pattern and its role in lymphomagenesis.

CONCLUSION AND FUTURE PERSPECTIVE
In conclusion, the pathogenetic role of miRNAs methylation in lymphomagenesis is just beginning to be unraveled. In contrast, there is considerable data to show that inactivation of tumor suppressor genes by DNA methylation plays important roles in the pathogenesis and prognostification of hematological malignancies (Chim et al., 2001a(Chim et al., ,b, 2003b(Chim et al., , 2004a(Chim et al., ,b, 2005a. For instance, methylation of retinoic acid receptor alpha (RARA), PTPN6, and DAPK genes has been implicated in pathogenesis of AML, myeloma, and CLL respectively (Reik and Walter, 2001;Chim et al., 2002Chim et al., , 2006Kaneda et al., 2004). Moreover, methylation of CDKN2B and WNT inhibitory factor 1 (WIF1) has been shown to be an independent prognostic factor predicting inferior diseasefree survival in acute promyelocytic leukemia (Chim et al., 2001b(Chim et al., , 2003b(Chim et al., , 2005a. Therefore, future studies will help to confirm the pathogenetic and prognostic impact of miRNA methylation in NHLs. DNA methylation can be reversed by demethylating agents 5-azacytidine or decitabine, and therapeutic benefits have been demonstrated in patients with myelodysplastic syndrome (Garcia-Manero and Fenaux, 2011). Besides, histone deacetylase inhibitors including vorinostat and panobinostat (Duvic and Vu, 2007), have been shown to be effective for some forms of lymphoma such as mycosis fungoides. As 5-azacytidine or decitabine may act synergistically with histone deacetylase inhibitors to enhance re-expression of genes silenced by hypermethylation, this strategy Frontiers in Genetics | Non-Coding RNA might also be applied to miRNA hypermethylation. Finally, there is hope that downregulated tumor suppressor miRNA in cancer may be restored by the use of miRNA mimics (Henry et al., 2011), which can form a partially double-stranded RNA mimicking endogenous pre-miRNA to be processed into an active miRNA molecule (Calin and Croce, 2009). miRNA can be delivered to the tumor in the form of an oligonucleotide mimic or by expressing the miRNA in the cancer using a gene vector (Henry et al., 2011).