Charles H. Lawrie, Nuffield Department of Clinical Laboratory Sciences, University of Oxford, Level 4, John Radcliffe Hospital, Oxford OX3 9DU, UK. E-mail: email@example.com
MicroRNAs are a recently discovered class of small (∼22nt) endogenously expressed translational-repressor RNAs that play key roles in many cellular pathways and whose aberrant expression appears to be a common feature of malignancy. MicroRNAs are expressed in specific haematological cell types and play important regulatory roles in early haematopoietic differentiation, erythropoiesis, granulocytosis, megakaryocytosis and lymphoid development. Additionally, there is an emerging body of research to suggest that microRNAs play an important role in the pathology of haematological malignancies. MicroRNAs have been found to act as both tumour suppressor molecules [e.g. MIRN15A (miR-15a), MIRN16-1 (miR-16-1)] in leukaemias and have oncogenic properties [e.g. MIRN155 (miR-155) and MIRN17–92 (miR-17–miR-92) cluster] in lymphomas. This review discusses the rapidly accumulating research that points to the major role microRNAs play in both haematopoiesis and haematological malignancy.
The first described microRNA, lin-4 was cloned and characterised as a translational repressor of developmental timing from Caenorhabditis. elegans by Lee et al (1993) and Wightman et al (1993). The transcript of this gene was highly unusual as it was non-coding, and produced extremely small transcripts (∼22nt) from hairpin structured RNA precursors. This remained something of an anomaly until 2000 when a second microRNA, let-7 was also cloned from C. elegans (Reinhart et al, 2000). Unlike lin-4, the sequence of let-7 was highly conserved in almost all organisms (Pasquinelli et al, 2000). It was soon realised that similar sequences were widespread in the genomes of eukaryotes that were first coined microRNAs by Lee and Ambros (2001).
Since then, the field of microRNAs has exploded and the number of publications dealing with this subject has increased almost exponentially year-on-year (Fig 1) with nearly 1500 citations to date (source PubMed database (http://www.pubmed.gov)). Thousands of microRNAs have now been described in a wide range of organisms including arthropods, nematodes, platyhelminthes, vertebrates, plants and viruses. There are currently 474 human cloned and characterised microRNA sequences deposited in the miRBase database (http://microrna.sanger.ac.uk/sequences/), although computational approaches suggest that the true figure is closer to one thousand (Bentwich et al, 2005; Berezikov et al, 2005).
MicroRNAs primarily function as translational repressors by binding to complementary target sequences in the 3′UTR (un-translated region) of mRNA. Despite the relatively small numbers of microRNAs, because a single microRNA can target several hundred genes, albeit targets almost exclusively predicted computationally, it is currently believed that between 10–30% of all human genes are a target for microRNA regulation (John et al, 2004; Lewis et al, 2005). In addition, a single target gene often contains putative binding sites for multiple microRNAs that can bind cooperatively (Lewis et al, 2003), allowing microRNAs to form complex regulatory control networks. The potential breadth of microRNA effects and indeed biological importance, is perhaps best illustrated by Lim et al (2005) who expressed MIRN124 (miR-124) and MIRN1 (miR-1) in HeLa cells and saw that the gene expression profiles of transfected cells shifted towards that typically seen in brain and skeletal muscle, respectively, the organs where these miRNAs are preferentially expressed.
Perhaps unsurprisingly then, microRNAs have been found to play key regulatory roles in a diverse range of pathways including control of haematopoiesis, developmental timing, cell differentiation, apoptosis, cell proliferation and organ development [reviewed by (Kim, 2005)] as well as in cancer, infectious disease (including virally encoded microRNAs), genetic disorders (Lin et al, 2006) and even heart disease (van Rooij et al, 2006).
microRNA biosynthesis and function
The majority of human microRNAs are encoded within introns of coding or non-coding mRNAs whilst others are located exgenically, within the exons of non-coding mRNAs or within the 3′UTR sequence of mRNA (Rodriguez et al, 2004). MicroRNAs are transcribed in a Poly II-dependent manner as 5′-capped large polyadenylated pri-microRNAs. Approximately 40% of human microRNAs are co-transcribed as clusters encoding up to eight distinct microRNA sequences in a single pri-microRNA transcript (Altuvia et al, 2005; Hertel et al, 2006). Pri-microRNAs are cleaved within the nucleus by Drosha, an RNaseIII-type nuclease, to form pre-microRNA 60–70 nucleotide hairpin structures (Fig 2). Drosha by itself possesses little enzymatic activity and requires the cofactor DiGeorge syndrome critical region 8 gene (DGCR8) in humans (Pasha in Drosophila) to form the microprocessor complex (Yeom et al, 2006). Once produced, the pre-microRNAs are exported from the nucleus to the cytoplasm by Exportin5 in a Ran-GTP dependent manner (Zeng, 2006). The cytoplasmic pre-microRNA is further cleaved to form an asymmetric duplex intermediate (microRNA: microRNA*) by Dicer, another RNaseIII-type enzyme. Similar to Drosha, cofactors such as TRBP and PACT (in humans) are necessary for Dicer activity (Lee et al, 2006). The microRNA: microRNA* duplex is in turn loaded into the miRNA-induced silencing complex miRISC in which Argonaut (Ago) proteins appear to be the key effector molecules. The strand that becomes the active mature microRNA appears to be dependent upon which has the lowest free energy 5′ end (Khvorova et al, 2003; Schwarz et al, 2003).
The consequence of miRISC-loaded microRNAs is largely dependent upon the degree of complimentarity between the microRNA and its target gene. Most plant microRNAs have perfect homology to their targets and their primary action appears to be the degradation of mRNA in a process similar if not identical to that found in the siRNA pathway. Although some animal microRNAs are believed to act in a similar fashion (Yekta et al, 2004), nearly all animal microRNAs contain only limited sequence homology to their targets, usually limited to a 5′ 6–8 nt ‘seed region’, and act in a more subtle way to repress translation without mRNA degradation. How this occurs remains unclear. Early experiments with C. elegans suggested that repression occurs by prevention of elongation or degradation of nascent translation products (Olsen & Ambros, 1999). It has since been suggested that mRNA bound to the microRNA–miRISC complex may be sequestered away from the translational machinery in P-bodies that additionally act in concert with enzymes to remove the 5′-cap from mRNA, hence preventing translation (Liu et al, 2005; Sen & Blau, 2005). Alternatively it has been suggested that microRNAs may prevent recognition of the 5′cap by translation factors (Pillai et al, 2005).
Aberrant expression of microRNA is a common feature of cancer
The potential importance of microRNAs in cancer is implied by the finding that the majority of human microRNAs are located at cancer-associated genomic regions (Calin et al, 2004a), and there is emerging evidence to suggest that dysfunctional expression of microRNAs is a common feature of malignancy (Calin & Croce, 2006). Moreover it has been suggested that microRNA expression profiling can distinguish cancers according to diagnosis and developmental stage of the tumour to a greater degree of accuracy than traditional gene expression analysis (Lu et al, 2005).
MicroRNAs are proposed to play a direct role in oncogenesis as they can function as both oncogenes (e.g. MIRN155 and members of MIRN17–92 cluster) and tumour suppressor molecules [e.g. MIRN15A (miR-15a) and MIRN16-1 (miR-16-1)] (see below). Aberrant expression of specific microRNAs has now been associated with many types of cancer including solid and haematopoietic tumours (Table I). The cause of aberrant expression of specific microRNAs in cancer can in some instances be explained by associated chromosomal abnormalities, encoding either the microRNAs themselves or components of the biosynthetic pathway (Zhang et al, 2006), but generally the underlying mechanism remains unknown, largely due to a lack of understanding of how microRNA expression is regulated.
Table I. Summary of microRNA involvement in cancer with emphasis on haematological malignancies.
The role of microRNAs in haematopoiesis or indeed in haematological malignancy has been little studied. Chen et al (2004) cloned approximately 150 microRNAs from murine bone marrow and found that MIRN181A (miR-181a), MIRN223 (miR-223) and MIRN142 (miR-142) were preferentially expressed in haematopoietic tissue. MIRN181A, found to be highly expressed in B-cell lineage samples, when expressed ectopically, caused a dramatic increase in the proportion of B-lineage cells produced. Although MIRN223 and MIRN142 were most highly expressed in myeloid lineages (the latter also in B-lineage cells), ectopic expression of these two microRNAs resulted in an increase in the proportion of T-lineage cells (30–50%) but not B-lineage or myeloid cells. Expression of the same microRNAs in human haematopoietic cells however, showed differing expression patterns from their murine counterparts (Ramkissoon et al, 2005). A further study showed that the microRNA expression profile varied significantly throughout murine haematopoietic development, specifically MIRN150 (miR-150) was found to be downregulated in response to T-cell stimulation by both Th1 or Th2 subsets whereas MIRN146 (miR-146) was upregulated only in Th1 cells (Monticelli et al, 2005).
The role of microRNAs in erythropoiesis was addressed by Felli et al (2005), who identified MIRN221 (miR-221) and MIRN222 (miR-222) as being highly expressed in human cord-blood derived haematopoietic CD34+ progenitor cells. Levels of these microRNAs were found to be downregulated in response to unilineage erythroid differentiation. Bioinformatic analysis suggested that both closely related microRNAs targeted KIT and levels of this mRNA were found to inversely correlate with levels of MIRN221 and MIRN222. Furthermore ectopic expression of these microRNAs in erythroid culture of CD34+ cells or the erythroidleukaemia cell line TF1 slowed cell growth and caused reduced levels of c-kit protein (Felli et al, 2005).
The function of MIRN223 in myeloid differentiation was studied by Fazi et al (2005) who demonstrated that granulocytic differentiation was regulated by MIRN223 in association with two transcription factors, NFIA and CEBPA. They found that NFIA and CEBPA compete for binding sites in the promoter sequence of MIRN223; NFI-A maintaining MIRN223 expression at low levels whilst CEBPA binding following retinoic acid-induced differentiation upregulates MIRN223 expression. Moreover, MIRN223 was demonstrated to inhibit translation of NFIA, resulting in a negative-feedback loop that favoured granulocytic differentiation. Expression analysis of in vitro-differentiated megakaryocytes derived from CD34+ bone marrow cells identified the downregulation of MIRN10A (miR-10a), MIRN10B (miR-10b), MIRN126 (miR-126), MIRN17 (miR-17) and MIRN20 (miR-20) as a feature of megakaryocytopoiesis (Garzon et al, 2006).
Recently a study of CD34+ haematopoietic stem-progenitor cells (HSPC) identified 33 microRNAs whose expression were common to HSPC derived from both bone marrow and peripheral blood including MIRN155, MIRN181A, MIRN221, MIRN222 and MIRN223, MIRN146 and MIRN16-1. Based on a bioinformatic approach to predict target genes for these microRNAs and functional assays to validate 18 haematopoietic differentiation-associated genes, the authors proposed that MIRN17, MIRN24 (miR-24), MIRN146, MIRN155, MIRN128 (miR-128) and MIRN181 may act to prevent the differentiation of early stage progenitor cells to a more mature stage whilst MIRN16-1, MIRN103 (miR-103) and MIRN107 (miR-107) may block differentiation at a later stage in HSPC development. MIRN221, MIRN222 and MIRN223 were proposed to control the terminal stages of haematopoietic development. Lentiviral expression of MIRN155 in K562 cells treated to induce differentiation into either myeloid and erythroid pathways resulted in greatly reduced colony formation, leading to the suggestion that MIRN155 controls myeloid and erythroid differentiation (Georgantas et al, 2007).
In addition, to the role of specific microRNAs, the general necessity of microRNAs in haematopoiesis has been evaluated in a murine system through the selective deletion of Dicer in the thymus using both a deletion at an early stage of T-cell development and at a much later stage (Cobb et al, 2005; Muljo et al, 2005). Surprisingly, in both experiments, this deficiency did not inhibit T-cell differentiation but did cause a severe block in peripheral CD8+ development and reduced numbers of CD4+ cells, which, when stimulated, underwent increased apoptosis and proliferated poorly. The CD4+ cells preferentially expressed IFN-γ suggesting a Th1 phenotype. It should be noted however, that the half-life of microRNAs in these experiments may be longer than expected or that microRNAs in other accessory cells may be important in T cell development. Thus, although Dicer appears to be essential for embryonic development (Bernstein et al., 2003; Kanellopoulou et al., 2005), the role of microRNAs in determining the later stages of lymphocyte development remains unclear.
MicroRNA expression in leukaemia
MicroRNA expression profiling of 38 patients with chronic lymphocytic leukaemia (CLL) demonstrated differences in microRNA expression that was associated with karyotype, the presence of ZAP-70 and IgVH mutation status (Calin et al, 2004b). Of particular interest, expression levels of MIRN15A and MIRN16-1, encoded within the 13q14 region, a deletion found in more than 65% of CLL cases, were downregulated in 75% of CLL cases that harboured this chromosomal abnormality. These microRNAs were subsequently shown to target BCL2 and to induce apoptosis in vitro, suggesting they have tumour-suppressor role in CLL (Cimmino et al, 2005). Consistent with this role, it has recently been demonstrated that MIRN16-1 negatively regulates cellular growth and cell cycle progression (Linsley et al, 2007). A follow-up study (Calin et al, 2005) of 94 CLL cases, defined a prognostically significant 13-gene microRNA signature by expression profiling. Moreover two of the CLL patients were found to have germline mutations in the MIRN16-1/MIRN15A precursor sequence that resulted in reduced expression levels of these microRNAs both in vitro and in vivo. Interestingly a homologous cluster to MIRN16-1/MIRN15A is encoded at 3q25 (MIRN16-2/MIRN15B), a region not commonly deleted in CLL. This raises questions about the importance of these microRNAs in the pathogenesis of CLL, although it is possible that consequent residual expression of these microRNAs might explain the more favourable prognosis observed in CLL patients with 13q14 deletions compared to other common CLL chromosomal abnormalities such as 17p13 or 11q23. Consistent with this idea, it has recently been reported that CLL patients with a monoallelic 13q14 deletion had slower lymphocyte growth kinetics than those with a biallelic deletions (Pfeifer et al, 2007).
A further indication of the role that microRNAs could play in the pathogenesis of CLL was provided by Pekarsky et al (2006) who found that the TCL1 oncogene, previously shown to be linked with an aggressive CLL phenotype, was regulated by MIRN29 (miR-29) and MIRN181Bin vitro and that levels of these microRNAs inversely correlated with levels of TCL1 expression in clinical samples of CLL.
Consistent with previous findings (Fazi et al, 2005), it was recently reported that the microRNA expression profile of acute promyelocytic leukaemia (APL) patients (and cell lines) treated with retinoic acid resulted in the upregulation of several microRNAs including MIRN223 as well as MIRN15A, MIRN15B, MIRN16, MIRNLET7A (let-7), MIRN342 and MIRN107 whilst MIRN181B was shown to be downregulated (Garzon et al, 2007).
In addition, there have been isolated case studies of leukaemia patients with translocations involving microRNAs. Gauwerky et al (1989) identified multiple chromosomal aberrations in a patient with aggressive promyelocytic leukaemia including a t(8;17) translocation that fused the promoter and 5′ region of MIRN142 (originally identified as BCL3) to a truncated MYC gene. It was proposed that this unusual translocation may account for the progression of this normally indolent disease to the more aggressive form observed. A case of acute lymphoblastic leukaemia was found to contain a translocation involving the complete MIRN125B sequence fused to the IGH gene locus (Sonoki et al, 2005).
MIRN155 and the MIRN17–92 cluster: lymphoma-associated oncogenes
Unlike leukaemia, a systematic identification of lymphoma-associated microRNAs has not yet been undertaken. Instead, a more focused approach has identified MIRN155 and components of the MIRN17–MIRN92 cluster as being lymphoma-associated microRNAs with potential oncogenic activity.
The non-coding BIC locus was originally identified as a common retrovirus integration site for avian leukosis virus that, despite being poorly conserved between avian, mouse and human genomes, could enhance lymphogenesis in cells that also over-expressed MYC (Clurman & Hayward, 1989; Tam et al, 1997). Commenting on the observation by van den Berg et al (2003) that BIC was highly expressed in over 90% of Hodgkin lymphoma (HL) cases, Metzler et al (2004) proposed that a phylogenically conserved region of 138 nucleotides in the BIC gene encoded a functional precursor sequence of MIRN155. Subsequently, both BIC and MIRN155 transcript levels were found to be upregulated 10- to 30-fold in diffuse large B-cell lymphoma (DLBCL) cases with higher levels of expression observed in cases with an ABC-immunophenotype than those with a GCB-immunophenotype (Eis et al, 2005). Similar findings were reported by Kluiver et al (2005) who also found that BIC over-expression in HL cases correlated with miR-155 expression and that primary mediastinal B-cell lymphoma, but not other non-Hodgkin lymphomas, highly expressed BIC. Over-expression of MIRN155, in Burkitt lymphoma (BL) and post-transplantation lymphoproliferative disorder (PTLD) at least, appears to be associated with EBV latency type-III infections (Jiang et al, 2006; Kluiver et al, 2006a).
Transgenic mice carrying the MIRN155 precursor sequence under control of a VH promoter-Ig heavy chain E enhancer, which becomes active at the late pro-B cell stage of B-cell development, were found to develop initially pre-B cell proliferation in spleen and bone marrow followed by a frank B-cell malignancy resembling high grade lymphoma after 6 months (Costinean et al, 2006). It should be noted however that these mice developed a polyclonal lymphoproliferation, suggesting additional factors are necessary for oncogenesis in this model.
In addition to its proposed role in haematopoiesis, MIRN155 was found to be upregulated during the macrophage inflammatory response (O'Connell et al, 2007). Furthermore, the authors demonstrated that inhibition of JNK resulted in downregulation of MIRN155 when stimulated by IFN-β, suggesting the involvement of the AP-1 transcription factor complex. The promoter region of BIC was previously reported to contain a putative AP-1 binding site (van den Berg et al, 2003). Consequently, it would be tempting to suggest that the high levels of MIRN155 and BIC observed in HL could be explained by constitutive expression of AP-1 that is a hallmark of this disease (Mathas et al, 2002).
Regulation mechanisms of MIRN155, and probably many microRNAs, are most likely to be cell-type restricted as it has recently been shown that FOXP3, a transcription factor highly restricted in its expression to CD4+/CD25+ regulatory T cells, was found to regulate expression of MIRN155 in these cells (Zheng et al, 2007). Consistent with this idea ectopic expression of the BIC gene in the Epstein–Barr virus (EBV)-negative BL cell line Ramos, but not in HEK293 or EBV type-III positive BL cell line Raji, resulted in a blockage of processing to form mature MIRN155 (Kluiver et al, 2006b).
A commonly found amplification in B-cell lymphomas, the 13q31 locus was shown to encode a functional precursor microRNA sequence, the MIRN17–92 cluster, that itself encodes seven mature microRNA sequences (He et al, 2005). Levels of the MIRN17–92 cluster were found to be elevated in four cell lines harbouring this amplification compared to five cell lines which did not. The same authors reported that 65% of 46 lymphoma cases (13 of which were DLBCL) over-expressed MIRN17–MIRN92, compared to 18/70 (26%) of cases of DLBCL that were shown to have 13q31 amplification by comparative genomic hybridisation (Karnan et al, 2004). Additionally, amplification of 13q31 was linked to over-expression of MIRN17–92 in lung cancer (Hayashita et al, 2005). Our data suggests that over-expression of these microRNAs is a more common occurrence in haematological malignancies than previously believed, as high expression levels were observed not only in all B cell lines tested, including those that are known to lack the 13q31 amplification (e.g. SUD-HL6, Raji and Karpas 422 (Ota et al, 2004)) but also T-cell lines and myeloid lines (Fig 3). In addition, over-expression of MIRN17–92 has been described as a common feature of solid tumours (Volinia et al, 2006). These data call into doubt the importance of 13q31 amplification as a mechanism for over-expression of the MIRN17–92 cluster in cancer.
Similar to BIC (Clurman & Hayward, 1989; Tam et al, 1997), He et al (2005) found that expression of components of the MIRN17–92 cluster in mice also over-expressing MYC greatly accelerated lymphogenesis. Compared to the Eμ-Myc mice, mice also over-expressing the MIRN17 cluster showed reduced levels of apoptosis, suggesting that the main effect of these microRNAs was to suppress cell death. Interestingly, increased lymphogenesis was only observed when these microRNAs were expressed together, but not as individual microRNAs, suggesting a cooperative effect. O'Donnell et al (2005) found that the MIRN17–92 cluster itself was upregulated through direct MYC binding. They also reported that components of the cluster, MIRN17-5-p and MIRN20 negatively regulated expression of E2F1, a pro-apoptotic transcription factor. Conversely, it was recently found that the promoter sequence of the MIRN17–92 cluster contains two functional E2F binding elements and chromatin immunoprecipitation analysis demonstrated that E2F3 binds this region, forming a regulatory loop between the pro-apoptotic E2F1 and the proliferative E2F3 (Woods et al, 2007).
Future directions of microRNA research
The rush for identifying microRNAs associated with cellular and disease mechanisms continues unabated. However, understanding the biological function of identified microRNAs is perhaps the biggest challenge facing the field at the moment. The major hurdle in associating biological function to a particular microRNA is the identification of its target genes. Currently by far the majority of target genes are predicted computationally and only a handful of genes have been validated experimentally. This is further compounded by the fact that different algorithms often predict non-overlapping targets for the same microRNAs (John et al, 2004; Krek et al, 2005; Lewis et al, 2005; Griffiths-Jones et al, 2006). There is clearly a need for a better understanding of microRNA: target interactions in order to advance the field, perhaps measuring the effect of exogenously adding microRNAs to microRNA-deficient models.
In summary, although it is clear that the functional importance of microRNAs in both haematopoiesis and haematological malignancies is gaining momentum rapidly, it is equally clear we still have much to learn from these tiny molecules.
This work was funded by grants from the Leukaemia Research Fund and the Julian Starmer-Smith Memorial Fund.