MicroRNAs regulate the expression of protein-coding genes in animals and plants. They function by binding to mRNA transcripts with complementary sequences and inhibit their expression. The level of sequence complementarity between the microRNA and mRNA transcript varies between animal and plant systems. Owing to this subtle difference, it was initially believed that animal and plant microRNAs act in different ways. Recent developments revealed that, although differences still remain in the two kingdoms, the differences are smaller than first thought. It is now clear that both animal and plant microRNAs mediate both translational repression of intact mRNAs and also cause mRNA degradation.
- seed sequence
- target recognition
- translational repression
miRNAs (microRNAs) are negative regulators of gene expression in both plants and animals . Mature miRNAs are produced by a multistep process (see Chapter 2) and are incorporated into a protein complex called the RISC (RNA-induced silencing complex). Since the mature miRNA in the RISC is single-stranded, it has the ability to anneal to another RNA molecule with a complementary sequence. miRNAs can therefore bind to specific target regions within mRNA transcripts which contain a sequence that can anneal to the miRNA. These sequences are called miRNA target sites and are usually at the 3′-UTR (untranslated region) in animals, but are in the coding region of plant mRNAs. As miRNAs bind target sites on mRNAs, they guide the RISC to these transcripts leading to the silencing of those messages. The present chapter focuses on the mechanism of this silencing.
One would expect that it is relatively straightforward to identify miRNA target sites if the sequence of all of the miRNAs and mRNAs are known. However, in animals and humans it is very rare that an mRNA contains a perfect complementary target site for any miRNA . The only known exception is the HOXB8 mRNA, which is recognized by miR-196 . All other human miRNAs silence mRNAs with target sites that are not perfectly complementary . Identifying these target sites by a similarity search is difficult because, by allowing mismatches between miRNAs and mRNAs in a database search, hundreds or thousands of potential target sites will be identified due to the small size of the miRNAs. Furthermore, the positions of the mismatches are important and, typically, nucleotides at positions 2–8 on the miRNA are perfectly complementary to the target mRNA . This region is called the seed sequence. On the basis of the number and position of mismatching nucleotides between the miRNA and its target site, three groups of target sites exist. The first group contains the 5′-dominant canonical target sites, which do not contain mismatches in the seed site and also show extensive base pairing to the rest of the miRNA. The second group is the 5′-dominant seed-only target sites, which are also perfectly complementary to the seed sequence, but only have limited base pairing with the rest of the miRNA . The third group contains mismatches in the seed site, but also show extensive base pairing with the 3′ half of the miRNA. Therefore the sites in this third group are called 3′-compensatory target sites . Several computational programs have been developed to predict miRNA target sites and because each of them uses a slightly different algorithm, the predicted target lists often show very little overlap with each other; e.g. TargetScan focuses on finding 5′-dominant canonical and seed-only target sites, whereas MiRanda preferentially predicts 3′-compensatory target sites .
It is much easier to predict miRNA targets computationally in plants because all known target sites show a very high complementarity to the entire miRNA. This does not necessarily mean that plant miRNAs cannot recognize target sites with several mismatches like in animals. However, because target sites with near perfect complementarity do exist, these have been extensively studied and little effort has been made to explore the potential targets with several mismatches.
Animal miRNAs repress translation
In order to understand how miRNAs repress translation, we first need to introduce a few features of mRNAs that are important for translation. Eukaryotic mRNAs are protected against non-specific exonuclease digestion by the cap structure at the 5′-end and by the polyA tail at the 3′-end. Efficient translation requires interaction of these two features, which leads to looping of the mRNA. The PABPC [cytoplasmic PABP (polyA-binding protein)], as the name suggests, is bound to the 3′-polyA tail, but also interacts with eIF4G (eukaryotic translation-initiation factor 4G). Because eIF4G also interacts with eIF4E, which recognizes and binds the 5′-cap structure, this sequence of interactions leads to circularization of the mRNA, which is required for efficient translation (see more on translation in ). Another feature of translation is the formation of polysomes. In the presence of translation inhibitors that arrest ribosomes (e.g. cycloheximide), mRNAs stay bound to ribosomes that have already initiated translation. Since more than one ribosome translates a single mRNA at once, mRNAs with various numbers of ribosomes can be purified on a sucrose gradient, these are called polysomes .
The first question in elucidating the mechanism of translational repression by miRNAs is to determine whether miRNAs either suppress the initiation of translation or act at the post-initiation stage. Interestingly, there is evidence for both initiation and post-initiation suppression; e.g. Petersen et al.  found that miRNAs which target mRNAs are present in the polysome fraction, although the protein encoded by those target mRNAs were not detectable. Since ribosomes already initiated translation of mRNAs present in the polysome fraction, it was concluded that these targets were silenced at the post-initiation stage. This model was further supported by the observation that miRNA silencing occurred in the absence of the 5′-cap structure . In this experiment, an IRES (internal ribosome entry site) was used instead of the canonical 5′-cap. However, other groups obtained contradictary results showing that target mRNAs were not in the polysome fraction, but are in the ribosome-free fraction . Furthermore, cap-independent translation initiated at IRESs was not silenced by miRNAs in these experiments. This suggested that miRNAs act at the initiation step and somehow inhibit ribosomes from binding to target mRNAs through a cap-dependent mechanism. This model gained further support from experiments that used cell-free extracts to study the miRNA silencing mechanism [9–13]. In these studies, miRNA silencing required an m7Gppp-cap and did not affect mRNAs with an artificial Appp-cap structure, or mRNAs with an IRES [12,13].
Animal miRNAs cause target degradation
The first mRNAs identified as miRNA targets were regulated by lin-4 in Caenorhabditis elegans. These mRNAs were not affected by lin-4 at the RNA level, but only at the protein level. Therefore it was initially thought that miRNAs in animals do not degrade mRNAs, but only repress translation. However, recent transcriptome studies showed that most miRNA targets are less abundant in the presence of miRNAs and the transcript level is also increased when miRNA activity is blocked [14–17]. When siRNAs (small interfering RNAs) with perfect complementarity to their targets (or plant miRNAs with near perfect complementarity; see below) are applied to cell cultures, the target cleavage happens at a specific position between nucleotides that are annealed to the 10th and 11th nucleotides of the miRNAs . However, animal and human miRNAs rarely cause cleavage at that position. Instead, a growing body of evidence indicates that miRNAs channel their targets to the cellular 5′-to-3′ mRNA decay pathway [19,20]. In this pathway mRNAs are first de-adenylated by the CAF1–CCR4–NOT deadenylase complex, followed by a DCP2 (mRNA decapping enzyme 2)-mediated de-capping (Figure 1). Loss of the 5′-cap structure allows XRN1, the major cytoplasmic 5′-to-3′ exonuclease, to degrade mRNAs (Figure 1). The evidence that miRNAs cause mRNA degradation through the cellular 5′-to-3′ mRNA decay pathway is shown by the increased level of miRNA target transcripts in the absence of components of this pathway [19,20]. Deadenylation of mRNAs by miRNAs was also detected in cell-free extracts [11,13]. The difference between results obtained using cell cultures and cell extracts is that, in vivo, the deadenylated mRNAs are also decapped and then degraded but, in vitro, decapping and degradation does not follow deadenylation. This raised the possibility that deadenylation alone could be sufficient to silence a target mRNA . Initially this was a controversial topic because deadenylation was shown to precede translational repression [11,13], but also to occur after it . However, two recent studies have shown that miRNAs suppress translation initiation first, which is then followed by deadenylation and mRNA decay [22,23].
Despite these unresolved questions, it appears that animal and human miRNAs do cause target degradation, although the mechanism is different from the cleavage mechanism by plant miRNAs or siRNAs in animal cells. Since mRNA silencing involves both translational repression and mRNA degradation, it is an interesting question as to which one is more widespread. In order to answer this question one has to profile both the mRNA accumulation (to measure mRNA degradation) and protein levels (to determine translational repression). Transcriptome profiling of cells where a specific miRNA has been suppressed or overexpressed can be done using microarrays or next-generation sequencing-based RNA-seq. The proteome profiling is more challenging, but is made possible by the sophisticated method of SILAC (stable isotope labelling by amino acids in cell culture) . In SILAC two populations of cells are cultured separately, one in medium containing normal amino acids and the other in medium which contains amino acids labelled with a heavy isotope. For example, the normal medium contains arginine with 12C and the ‘heavy medium’ contains arginine with 13C. Cells growing on the ‘heavy medium’ incorporate the 13C-containing amino acids, resulting in proteins in those cells being slightly heavier than proteins in cells grown in normal medium. After protein extraction, the samples can be mixed and analysed on the same gel. This is followed by mass spectrometry, which can distinguish between the lighter and heavier form of the same protein and can provide accurate information about the ratio of the two forms. Therefore if one sample contained a higher or lower level of a specific miRNA, its effect on the proteome can be measured in SILAC experiments . Two such studies have been carried out and both concluded that the extent of translation suppression by miRNAs is quite modest, usually the reduction in protein level was found to be less than 4-fold [14,15]. However, there are differences between the two studies' results concerning the dominance of one mechanism over the other. Baek et al.  found that very few mRNAs were repressed translationally without degradation and those targets showed weak silencing. However, Selbach et al.  found that at an early time point, many targets showed reduced protein levels, but unchanged mRNA levels, although most targets showed reduced mRNA and protein levels at a later time point. These differences demonstrate that more work is necessary to fully understand the extent and distribution of translational suppression and RNA degradation with respect to miRNA function. The two groups used different cell types and analysed the effects of different miRNAs so the results are not directly comparable. It is conceivable that different miRNAs could act principally through one of the pathways, and then change to the other during development or perhaps in different tissue types.
Plant miRNAs repress translation
The first two examples of translational repression of target mRNAs by plant miRNAs were the miR-172-targeted APETALA2 and miR-156/157-targeted SPL3 [25,26]. On the other hand, there were many examples of miRNA-mediated target degradation, implying that these two cases were exceptions. However, a forward genetic screen unexpectedly revealed that many plant miRNAs potentially repress translation of target mRNAs in addition to degradation . The Arabidopsis plant that was mutated contained a transgene coding for the GFP (green fluorescent protein) with an engineered miR-171 target site in its 3′-UTR. Wild-type transgenic plants expressed the GFP protein at a low level because the endogenous miR-171 targeted the transgenic GFP mRNA. After mutagenesis, the seedlings were screened for increased GFP expression and two types of mutants were identified: ones that were miRNA biogenesis deficient (mbd) and ones that were miRNA action deficient (mad). Two mbd mutants were found where miRNAs accumulated at a reduced level and these turned out to be new alleles of DCL1 (dcl1-12) and HEN1 (hen1-7). All six mad mutants on the other hand contained normal levels of miRNAs and could be grouped into two classes. In the first class (mad1–4), target mRNAs accumulated at a higher level leading to increased GFP protein levels. More interestingly, the other two mad mutants (mad5 and mad6) contained a low level of GFP mRNA similar to the wild-type plants, but an increased level of GFP protein. The authors ruled out that it was a specific effect on GFP mRNA by showing that GFP mRNA without the miR-171 target site was not affected in the mutants. They also demonstrated that some endogenous target genes, which were regulated by various miRNAs, behaved similarly to the GFP reporter gene (where the miRNA target sites were either in the 5′-UTR, coding region or 3′-UTR). On the basis of these results it was concluded that plant miRNAs can either degrade or repress the translation of intact target mRNAs, as mad1–4 were required for degradation and mad5 and mad6 were necessary for translational repression. This discovery raised several questions about plant miRNAs. Although mad5 was identified as KTN1, which encodes the p60 subunit of the microtubule-severing enzyme katenin, the molecular mechanism of miRNA-mediated translational repression remains to be described. Also, mad5 and mad6 do not display strong phenotypic differences compared with wild-type plants. Considering that a reduced level of miRNAs in various biogenesis mutants leads to strong developmental defects, the importance and extent of the translational repression needs to be clarified. Nonetheless, it is clear that plant miRNAs can repress translation, which was further supported by a biochemical study showing that miRNAs and Ago1 can associate with polysomes .
Plant miRNAs cleave target mRNAs
As mentioned previously, most plant miRNAs show near perfect complementarity to their targets and cause endonucleolytic mRNA cleavage . This so called ‘slicing’ is carried out by Ago1 and happens at the specific position between the nucleotides that are complementary to the 10th and 11th nucleotides of the miRNA (Figure 2). The cleavage generates two fragments: the 5′ fragment is protected at its 5′-end by the cap structure, but has an unprotected 3′-end, and the 3′ fragment has an exposed 5′-end and a polyA-protected 3′-end. The 5′ fragment is processed from the 3′-end by the exosome and the 3′ fragment is degraded by XRN4 . Although both fragments are degraded to some extent, most 3′ fragments are more stable than their 5′ counterparts, and some of them can even be detected by Northern blot analysis . Since the 3′ fragments are relatively stable and contain a monophosphate at their 5′-end, an adapter can be ligated to their 5′-end and 5′-RACE (rapid amplification of cDNA ends) analysis can be carried out to determine the exact 5′-end sequence. If the cleavage occurs exactly at the expected position (between the 10th and 11th nucleotides) of a predicted target site, this is usually accepted as an experimental proof for target validation. On the basis of this approach, a genome-wide target identification technique was developed called either degradome library sequencing  or PARE (parallel analysis of RNA ends) . In the first step total RNA samples are enriched for polyA-containing mRNAs and an adapter is ligated to the 5′-end. Then the first strand of cDNA is synthesized using an oligo(dT) primer followed by a PCR using primers that can anneal to the adapter and the oligo(dT) primer. The adapter contains a restriction site which is recognized by a restriction enzyme that cleaves away from the recognition site, exactly 21 nt downstream of the original 5′-end. After digestion with the restriction enzyme, a dsDNA (double-stranded DNA) adapter is ligated to the fragment, therefore the final product contains the 5′ adapter, the first 21 bp of the cDNA from the 3′ cleavage fragment and the 3′ adapter. Since both adaptors are compatible with the Illumina platform, the ligation products can be sequenced by a high-throughput method yielding tens of millions of sequence tags. The sequencing reads are mapped to the genome and then tested as to whether they are the 3′ halves of potential miRNA target sites and whether the 5′-end of the cleavage fragment is exactly at the expected position. The first degradome/PARE studies used the CleaveLand software  to analyse the sequencing results, which can identify sequence tags that map to predicted target sites of miRNAs that were given as inputs (i.e. it requires prior knowledge of small RNAs that would potentially cleave mRNAs). Using this tool normally 100–150 target mRNAs are identified for known miRNAs [31–33]. Recently a new tool was developed (PAREsnip), which does not require information about small RNAs that could potentially cleave target mRNAs . PAREsnip is able to use an entire small RNA library containing tens of millions of reads to search for sequence tags mapping to cleavage sites by any small RNA. Studying Arabidopsis degradome libraries identified more than 4000 sites that fulfilled the strict criteria of miRNA–target site interactions and were found in two independent degradome libraries . Since this number is substantially higher than the number of targets found for known miRNAs, it suggests that either there are many more unidentified miRNAs, or there are other classes of small RNAs that can cause mRNA cleavage, similar to miRNAs.
Over the last few years, huge advances have been made in elucidating the mechanism of miRNA-mediated mRNA targeting. Although miRNAs were initially thought to have distinct modes of function in animal and plant systems, this dogma has been revised following recent research developments. It is now clear that both mRNA degradation and translation repression occur in both kingdoms. However, the extent and occurrence of these two mechanisms has yet to be determined. There are still clear differences between plants and animals as the degradation of target protein-coding transcripts occurs through different modes-of-action. While plant miRNA complexes have slicing activity, animal miRNA complexes mediate degradation through the XRN pathway; however, plants have been shown to also use the XRN pathway to process sliced fragments in addition to the exosome. It is quite possible that both target cleavage and translational repression are used to some degree by the cell, perhaps depending on the developmental stage of the tissue in which they reside, the tissue type or the miRNA in question.
• MicroRNAs regulate the expression of protein-coding genes in animals and plants.
• Plant microRNAs cause target mRNA cleavage at a specific position, opposite to the 10th and 11th nucleotides of the microRNA.
• Plant microRNAs also cause translational suppression.
• Animal microRNAs do not cause cleavage at a specific position, but they trigger decapping and deadenylation leading to mRNA decay.
• Animal microRNAs also cause translational suppression.
- © The Authors Journal compilation © 2013 Biochemical Society