Tamas Dalmay, School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK. (fax: +0044 1603 592250; e-mail: firstname.lastname@example.org).
The Nobel Prize in Medicine and Physiology was awarded to the RNA interference (RNAi) field in 2006 because of the huge therapeutic potential this technique harbours. However, the RNAi technology is based on a natural mechanism that utilizes short noncoding RNA molecules (microRNAs) to regulate gene expression. This paper reviews our current knowledge about microRNAs focusing on their involvement in cancer and their potential as diagnostics and therapeutics.
The human body consists of hundreds of different tissues. The cells in different tissues carry out specific functions although all cells in the body contain identical genetic information on the chromosomes. There are approximately 30 000 genes in our chromosomes but not all genes are active in a cell. Cells are different from each other, in spite of having the same genes, because only a set of genes are expressed in each cells determining the characteristics and function of the cells and tissues made up of those cells. A gene is active or expressed if messenger RNA (mRNA) is produced from it in the nucleus that is transported into the cytoplasm for translation into proteins. It is clear that precise regulation of gene expression (how much protein is produced from a gene at a certain time and place) is crucial for normal development and function of the body. Several layers of gene expression regulation were discovered in the last 50 years including switching on and off transcription (transcriptional), mRNA stability (post-transcriptional), efficiency of protein production (translational) and stability of proteins (post-translational). Several players are known in these processes such as proteins (transcription factors or suppressors) that can boost or inhibit gene expression, DNA sequences that are localized around genes and are recognized by proteins (promoters, terminators, enhancers and other cis-acting elements) and other proteins that can destabilize mRNAs or proteins. However, RNA was not implicated in gene expression regulation. The only appreciated role of RNA was to help making protein from DNA either by providing the template for translation (mRNA) or as accessory tools for translation (ribosomal RNA and transfer RNA).
After determining the nucleotide sequences of several genomes (the whole DNA content of cells in an organism) such as worm, fruit fly, zebrafish, chicken, mouse and human, researchers realized that only a small percentage of the genome is made up by genes and most of the genome do not code for proteins. Initially the regions that were outside of genes were called junk DNA. However, later it was realized that RNA is transcribed from many intergenic regions (DNA stretches on the chromosomes between genes) and these RNA molecules were called noncoding RNA (ncRNA). The function of many ncRNAs is still unknown but our understanding of one specific group of ncRNAs has grown significantly in the last 5 years. This group of ncRNAs is called microRNAs (miRNAs).
As the name suggests miRNAs are very small RNA molecules, only 21 nucleotides (nts) in length. To put it into context, the size of most mRNAs is between 2000 and 5000 nts; therefore it is not a surprise that miRNAs were remained undiscovered until recently. Several breakthroughs, some of them accidents, led to the discovery of miRNAs. The first miRNA, although it was called differently as short temporal RNA (stRNA) because it was expressed temporally, was identified by Lee et al.  through a mutant Caenorhabditis elegans (roundworm). The mutant worm was stuck at a certain developmental stage and map-based cloning identified the lin-4 gene responsible for this phenotype. Lin-4 encoded for an 80 nts noncoding RNA, however, at that time it was considered as an oddity of C. elegans genetics because lin-4 was not present in other species. Another stRNA (let-7) was later found also in C. elegans by Reinhart et al.  but let-7 had homologous sequences in several other species including human suggesting that stRNAs are not specific to worms.
The other line of research leading to miRNAs helped us to understand their function. It was observed in both animals and plants that in certain artificial situations (such as transgenic plants and worms injected with RNA) the expression of genes was silenced in a sequence specific way. Protein was not produced from the silenced gene although mRNA production was the same as in nonsilenced individuals. Fire et al.  showed that long double stranded RNA (dsRNA) was responsible for the sequence specific gene silencing (what they called RNA interference or RNAi) and Fire and Mello received the Nobel Prize for this discovery in 2006. On the other hand, Hamilton and Baulcombe  found that gene silencing in plants was correlated with the presence of a 21 base pair (bp) dsRNA species, called short interfering RNA (siRNA). Soon after these discoveries it was found that long dsRNA molecules are processed into siRNAs by an enzyme called Dicer both in animals and plants . However, cells do not normally contain long dsRNAs therefore this activity of Dicer was not considered its main function. This was confirmed by the finding that Dicer (that is present in most species) was also responsible for the production of lin-4 and let-7 in C. elegans and in fact let-7 was also produced in a Dicer dependant manner in human HeLa cells . These lines of research pointed to the conclusion that short temporal RNAs should exist in most species and their function could be similar to siRNAs that is to target mRNAs and suppress their activity. The final proof came when three groups identified many 21 nt RNA from different species by direct cloning (generating cDNA from the short RNAs and sequencing the cDNA library) [7–9]. At this point stRNAs were renamed as miRNAs because some of them were expressed all the time instead of temporally.
Biogenesis and activity of miRNAs
microRNAs genes are transcribed by RNA polymerase II just like protein coding genes . The primary transcript (pri-miRNA) is capped at the 5′ end and polyadenylated at the 3′ end similarly to mRNAs  (Fig. 1). The length of pri-miRNAs is not known but often it contains several mature miRNAs (see later) that form clusters. Pri-miRNAs are cleaved in the nucleus by an Rnase III type enzyme Drosha  and the product of this cleavage is one or more (in case of clusters) precursor miRNA (pre-miRNA). Pre-miRNAs are usually 70–90 nt long molecules with a strong secondary structure forming a stem and a loop (hairpin structure; Fig. 2). Pre-miRNAs are transported to the cytoplasm where they are further processed by Dicer that recognizes the stem as dsRNA and releases a 21 bp dsRNA (miRNA duplex). One of the strands of this duplex is incorporated into a protein complex called RNA induced silencing complex (RISC; . The strand incorporated into RISC is the mature miRNA whilst the other strand (called miRNA* strand) gets degraded. The mature miRNA guides RISC to mRNAs containing a partially complementary sequence (miRNA target site) to the miRNA. If the miRNA can anneal to the target site, RISC cleaves the mRNA (if the complementarity is near perfect), degrades the mRNA through de-capping and/or de-adenylation or suppresses the translation without affecting the mRNA level (Fig. 3). Therefore miRNAs represent a new layer in gene expression regulation: short RNA molecules encoded by the genome that regulate the expression of protein coding genes ensuring normal development and functioning of the body.
MicroRNAs in development and differentiation
Since lin-4 and let-7 was discovered through developmental screens in C. elegans it is not surprising that other miRNAs are also involved in development of higher organisms including mammals (Table 1). MiRNAs seem to have a role during early stages of development suggested by the presence of a set of stem cell specific miRNAs that are downregulated during differentiation [14, 15]. Another example of miRNA activity at an early stage is the targeting of Hox genes (involved in developmental patterning) by miR-196 . During embryogenesis many miRNAs show tissue specific expression pattern, which was demonstrated by in situ hybridization experiments [17–19]. The muscle specific miRNAs miR-1 and miR-133 are transcribed together from the same cluster but they have opposing roles. It was suggested that miR-1 promoted differentiation by targeting histone deacetylase 4 (HDAC4), a transcriptional repressor of muscle gene expression, whilst miR-133 suppressed differentiation by repressing serum response factor . Transgenic mice over-expressing miR-1 display premature differentiation of cardiomyocytes because of a decreased pool of proliferating ventricular cardiomyocytes causing heart failure . Deletion one of the miR-1 copies (miR-1-2) in mice led to a pleiotropic effect including altered cardiac morphogenesis, electrical conduction and cell-cycle control . Another function of miR-1 was discovered by a different approach. A particular type of sheep (Texel) exhibit increased muscle formation and a quantitative trait locus was mapped to identify this economically important feature. The locus contains the myostatin gene that is a negative regulator of skeletal muscle mass. This gene carries a mutation in Texel sheep creating a target site for miR-1. As miR-1 is highly expressed in muscle it can efficiently down-regulate myostatin expression in Texel sheep leading to muscle mass increase .
Table 1. Roles of microRNAs (miRNAs) in development
Several miRNAs are specifically expressed or enriched in the central nervous system (CNS). The most abundant CNS specific miRNA is miR-124, which shifted the mRNA expression profile of HeLa cells to that of brain cells after transfection of HeLa cells with the miRNA . However, inhibition or over-expression of miR-124 did not significantly change the acquisition of neuronal fate, suggesting that miR-124 is probably not a primary determinant of neuronal differentiation . The same authors showed that laminin gamma 1 and integrin beta 1 are suppressed by miR-124, both of which are highly expressed in neural progenitor cells but repressed upon neuronal differentiation . Similarly, phosphatase SCP1 (small C-terminal domain phosphatase 1) was shown to have anti-neural function during brain development. Down-regulation of SCP1 during CNS development is critical for neurogenesis and miR-124 contributes to this process by suppressing SCP1 expression . MiR-124 also targets PTBP1 (polypyrimidine tract binding protein 1), which encodes a splicing factor that produces an unstable form of PTBP2 mRNA in non-neuronal cells . During neuronal differentiation PTBP1 level is reduced by miR-124, leading to the accumulation of correctly spliced PTBP2 mRNA and a significant increase in PTBP2 protein. Other miRNAs also have important function in nervous system such as miR-34 that is involved in dendritic spine development  and miR133b that regulates the maturation and function of midbrain dopaminergic neurons by targeting the paired-like homeodomain transcription factor Pitx3.
Other examples for developmental roles of miRNAs include regulation of B-lymphocyte and myoblast differentiation by miR-181 [29, 30], insulin secretion by miR-375 , liver development and function by miR-122  and adipocyte differentiation by miR-143 .
miRNAs and cancer
The level of miRNAs and their targets are well balanced in healthy cells. Some examples are discussed above when altering the level of miRNAs (knock-out or over-expression) resulted in disease symptoms. One of the most serious diseases in the world is cancer, which is a complex disease. The development of cancer depends on the interaction of several factors. Oncogenes promote cell proliferation and tumourigenesis whilst tumour suppressor genes repress cell division and tumour formation. There are different oncogenes and tumour suppressor genes expressed in different tissues. The gene expression profile of tumours in different tissues is therefore different but it also changes during development of cancer. It is because cancer has five major stages (initiation, promotion, malignant conversion, progression and metastasis) and different genes are important in each step. In a simplified model the tumour suppressor genes control oncogenes and as deleterious mutations are accumulated in tumour suppressors (or mutations into oncogenes that boost their expression) either by ageing, because of carcinogens from the environment or through genetics, the balance tips towards oncogene activity and cancer develops. However, this is a very complicated process and the exact molecular mechanism of cancer formation is still not known.
As miRNAs affect gene expression they were good candidates for keeping the balance between tumour suppressors and oncogenes (Table 2). The first report that underlined this hypothesis came from Calin et al.  who were studying a region at chromosome 13q14 that is frequently deleted in chronic lymphocytic leukaemia (CLL). They found that the smallest region that is commonly missing from CLL patients contains two miRNA genes, mir-15a and miR16-1. These two miRNAs derive from the same primary transcript and down-regulation of this transcript is associated with the 13q14 deletion. The proposed tumour suppressor role of miR-15a and miR-16-1 was supported by the discovery of germ-line point mutations in two CLL patients that decreased the accumulation of the two miRNAs . A third study elucidated the mechanism of miR-15/16 tumour suppression by identifying their target gene. Both miR-15a and miR-16-2 recognize target sites on the 3′UTR of BCL2, an anti-apoptotic oncogene . These miRNAs control the level of BCL2 in normal cells but the expression of the oncogene is increased when the miRNAs are deleted leading to CLL.
Table 2. MicroRNAs involved in different types of cancers
miRNAs as oncogenes and tumour suppressors in different tissues
One of the most common cancers of adults in economically developed countries is lung cancer. Reduced expression of let-7 showed good correlation with shortened postoperative survival . In vitro experiments confirmed this observation; increased expression of let-7 in lung adenocarcinoma cell lines suppressed cell proliferation. Tumour suppressor activity of let-7 was elucidated by validating two predicted target genes. RAS and MYC, key oncogenes in lung cancer, contain several let-7 target sites and are directly regulated by let-7 . In addition, RAS and MYC expression negatively correlated with let-7 expression in lung tumour samples . However, miRNAs can also be oncogenes as the miRNA cluster miR-17-92 demonstrates. This cluster contains 14 similar miRNAs and present in three copies in the human genome . The expression of this cluster is strongly upregulated in lung cancer, especially in small-cell lung cancer that is the most aggressive form on lung cancer . In line with the oncogene activity of these miRNAs their predicted targets are PTEN and RB2, two known tumour suppressor genes . Although these targets have not been validated their over-expression in lung cancer was shown to be c-myc (an oncogene) dependent [42, 40].
Most thyroid cancers are papillary thyroid carcinoma (PTC). miRNA profiling showed that three miRNAs (miR-221, miR-222 and miR-146) accumulated at a much higher level in thyroid tumours than in matching healthy tissues . One of the predicted targets of these miRNAs is KIT that was generally down-regulated in the PTC samples .
microRNAs were identified in colon tissue and two of them (miR-143 and miR-145) showed reduced expression in different stages of colorectal neoplasia . Interestingly, the primary transcript of these two miRNAs was not changed in cancer samples, only the accumulation of the mature miRNAs suggesting that the altered level of mature miRNAs is due to differential processing in normal and cancer tissues .
Breast cancer may be less frequent in the overall adult population but it is the most important cancer amongst females. miRNA profile of clinical samples (76 neoplastic and 10 normal breast tissues) was established using microarrays containing probes against all known miRNAs. Several miRNAs (miR-125b, miR-145, miR-21 and miR155) showed lower accumulation in cancer samples . The same study also found that the lower expression of these miRNAs correlated with tumour stage, oestrogen and progesterone receptor expression, proliferation index and vascular invasion.
Human brain cancer is one of the most difficult to cure. Gliobastoma multiforme (GBM) is the most common form of brain cancer in human but its development is poorly understood. Expression profile of miRNAs was studied in glioblastoma samples and miR-221 showed higher whilst miR-181 showed lower accumulation in GBM samples than in normal tissues . Chan et al.  found miR-21 upregulated in GBM and experimental suppression of miR-21 activity led to increased caspase dependent cell death in cultured gliobastoma cells.
The BIC gene has been linked to several types of lymphomas (cancer of the lymphocytes, a type of white blood cell in the vertebrate immune system). However, the BIC gene does not encode for a protein and it was considered as a noncoding RNA with oncogene activity. BIC expression is usually higher in Hodgkin and Burkitt lymphoma than in normal lymphoid cells  but the mechanism of BIC induced lymphoma is poorly understood. Comparison of BIC sequences of several species revealed that the conserved region contains a miRNA gene (miR-155; . These observations inspired a study of miR-155 accumulation in B-cell-derived lymphomas, which found a consistently high level of miR-155 expression . A putative target gene of miR-155 is the transcription factor PU.1 that is necessary for B-cell differentiation.
miRNAs and metastasis
The so far described examples illustrate that miRNAs target genes that play a role in cell proliferation and apoptosis. However, malignant tumourigenesis has other characteristics than reduced cell death and enhanced cell division such as the ability to spread. This is due to changes in the interaction between tumour and surrounding tissues that allow sustained angiogenesis and metastasis. A bioinformatic analysis by Dalmay and Edwards  systematically searched for potential miRNA target sites on 3′ UTR sequences of genes that are involved in metastasis. Many of these genes were found to be potential miRNA targets. Expressions of miRNAs that may target these genes were then investigated using the available miRNA expression profiles of cancer samples. The differential expression of several miRNAs between healthy and cancer samples was in line with the predicted function of these miRNAs (either oncogenes or tumour suppressors). The predicted metastasis related genes that are potentially under miRNA regulations are lysyl oxidase, E-cadherin, integrin alpha-beta3, syndecan-1, hypoxia-inducible factor-1alpha, tissue inhibitors of metalloproteinase-3, adamalysin metalloproteinase-17, CXCR-4 and c-Met that underline the morphological changes associated with malignancy .
There are already experimental evidences for miRNA involvement in metastasis. Expression of two anti-angiogenesis genes, thrombospondin-1 (Tsp1) and connective tissue growth factor (CTGF) is reduced during angiogenesis and both genes are predicted targets of miRNAs in the miR-17-92 cluster. Dews et al.  validated the prediction by knocking down and over-expressing the miRNAs. Reduced miRNA activity resulted in higher expression of Tsp1 and CTGF and increased level of the miRNAs led to reduced-level gene expression . In another, recent study, Ma et al.  found that the expression of miR-10b was increased in metastatic breast cancer cells compared with healthy or nonmetastatic tumourigenic cells. They also investigated the cause of this upregulation and identified the transcription factor (Twist) that was responsible for the higher miRNA expression. The author went on and validated HOXD10 as a target of miR-10b and showed that a decrease in HOXD10 level resulted in higher level of RHOC, which stimulates cancer cell motility .
miRNAs as cancer diagnostics
Many independent studies on different tissues demonstrated that cancer cells have different miRNA profiles compared with normal cells suggesting that miRNA profiles could be used for diagnosis. Technical developments allow us to measure the accumulation level of all known miRNAs using different approaches: microarray, bead-based flow cytometry and quantitative PCR. It is remained to be seen whether any of these techniques can be adapted to a clinical setting. Bead-based flow cytometry is the cheapest and therefore has the highest chance to become a widespread method for miRNA profiling. Using this technique Lu et al.  demonstrated the power of miRNA profiling in cancer diagnosis. They established the miRNA profiles of 17 poorly differentiated tumours and 12 samples were correctly diagnosed based on miRNA signatures. The mRNA profiles of the samples were also generated but only one sample was correctly diagnosed using the mRNA data. Although a miRNA profile contains much less information than an mRNA profile (a few hundreds versus 35 000) the potential value may be higher because of the regulatory role of miRNAs.
Therapeutic potentials of miRNAs
Although our understanding of the miRNAs’ role in cancer is still very limited their therapeutic potentials have been already proposed. There are two possible approaches: blocking oncogenic miRNAs or over-expressing miRNAs with tumour suppressor activity. Theoretically, anti-miRNA oligonucleotides (AMOs) can be used to suppress miRNA activity if the AMOs can bind strongly to the miRNA and are stable enough in physiological conditions . AMOs have complementary sequences to miRNAs and contain several chemical modifications to achieve these goals. Two types of modifications were developed to attain strong binding: 2′-O-methylation of RNA nucleotides [32, 56] and locked nucleic acid (LNA) DNA nucleotides . These modifications are also used in the opposite approach where modified oligonucleotides are delivered into cells that under-express miRNAs with tumour suppressor activity. This approach showed promising results in cell culture  and needs to be tested in animal models. The great challenge for both approaches is the specific delivery of oligonucleotides into the tumour tissue. As these molecules are equally active in healthy and cancer cells their side effects must be minimized before they can be considered for clinical trials.