MicroRNA GO annotation manual

From GO Wiki
Jump to navigation Jump to search

Scope of the guidelines

These guidelines are intended to cover firstly, annotation of the protein components of the canonical mammalian miRNA processing pathway (Figure 1 and Winter et al. 2009) and secondly, the annotation of the role of miRNAs in gene silencing together with the targets of miRNA regulation. The focus is on negative regulation of gene expression, or gene silencing, by miRNA regulation via the 3’UTR of mRNAs, although other mechanisms of gene regulation are mentioned. Additionally, there is a section on positive regulation of transcription by miRNA, since there are an increasing number of papers now showing this mechanism.

These guidelines do not cover the annotation of proteins involved in gene silencing after the miRNA processing pathway, although these proteins are listed in Table 1 for information. The guidelines do, however, cover annotation of proteins that regulate the cellular levels of specific miRNAs (e.g. TGF-beta). Only experimental data demonstrating the role of a miRNA in a process/function through addition of an exogenous miRNA (IDA) or reduction of a miRNA (IMP) should be annotated. Co-incident expression data associated with a specific condition (IEP) should not be annotated, this information is available through other resources, such as ArrayExpress; expression alone does not confirm the involvement of a miRNA in a biological process.

Since there are some major differences between mammalian and plant miRNA biogenesis and action, there is also a section specifically describing the current status of plant miRNA biogenesis and action.

Background

The canonical miRNA processing pathway starts with the production of the primary miRNA transcript (pri-miRNA) by RNA polymerase II with subsequent cleavage of the pri-miRNA by the microprocessor complex DROSHA–DGCR8 in the nucleus. This cleavage results in a precursor hairpin, the pre-miRNA, which is exported from the nucleus by RAN:GTP:XPO5. Once in the cytoplasm, DICER1 ribonuclease, in complex with one of the double-stranded RNA-binding proteins, TARBP2 or PRKRA (PACT), cleaves the pre-miRNA hairpin to its mature double-stranded length of around 22 nucleotides. The RNA-induced silencing complex (RISC)-loading complex, comprising of DICER1, TARBP2 (or PRKRA) and one of the argonaute proteins (AGO1-4), loads the functional strand of the mature miRNA into the RISC and the passenger strand is degraded. DICER1 and TARBP2/PRKRA then dissociate leaving the mature RISC (miRISC) consisting of AGO1-4 and miRNA. The miRNA then guides the RISC to silence target mRNAs through mRNA cleavage, translational repression or deadenylation (Figure 1).

Annotation of the miRNA processing pathway

The recommended annotations for those proteins involved in miRNA processing are shown in the left panel of Figure 1. Additional notes on some of the components of this pathway are given below.



Figure 1. The canonical mammalian miRNA processing pathway. The proteins involved in the pathway of miRNA formation are shown together with the GO terms that they are expected to be associated with. The relevant protein complex IDs could also be annotated with these terms. See also (Winter et al. 2009). Protein names: DROSHA: Ribonuclease 3; DGCR8: Microprocessor complex subunit DGCR8; XPO-5: Exportin-5; RAN-GTP: GTP-binding nuclear protein Ran; DICER1: Endoribonuclease Dicer; TARBP2: RISC-loading complex subunit TARBP2; AGO: Argonaute. (From Huntley et al. RNA. 2016 May;22(5):667-76).

Note: A miRNA identifier (or a mRNA identifier for Argonaute) should not be included in the annotation extension field for any of the annotations to the core miRNA processing machinery. This is because this pathway is applicable to thousands of miRNAs (or mRNAs) that are involved in (or affected by) the gene silencing pathway, and so is not informative information.

DROSHA and DICER1

DROSHA and DICER1 are both ribonuclease enzymes that have the same activity but differ in the site of their cleavage. As such, both of these proteins can be annotated with “ribonuclease III activity” (GO:0004525). The definition of this term is “Catalysis of the endonucleolytic cleavage of RNA with 5'-phosphomonoesters and 3'-OH termini; makes two staggered cuts in both strands of dsRNA, leaving a 3' overhang of 2 nt.” Figure 2 depicts the cleavage site within the miRNA for both enzymes.


Figure 2. DICER and DROSHA cleavage sites.

RISC complex and the RISC-loading complex

It should be noted that, in some older papers, authors did not distinguish between the RISC complex and the RISC-loading complex. The mature RISC complex now is known to be composed of an Argonaute protein (AGO1-4) and a miRNA, whereas the RISC-loading complex is composed of an Argonaute protein (AGO1-4), DICER1, and TARBP2 or PRKRA together with the miRNA. Figure 3 shows the distinction between the RISC-loading components and those of the RISC complex.

Figure 3. RISC assembly in human cells. a. The first step is RISC loading, in this step the miRNA/miRNA* duplex is transferred from DICER1 to AGO in the RISC loading complex (RLC) followed by dissociation of DICER1 and TARBP2. b. Next, the N domain of AGO actively wedges between the miRNA strands and c. the PAZ domain of AGO unwinds the miRNA duplex. d. The passenger strand (miRNA*) dissociates from the RISC and undergoes rapid degradation, by a mechanism that is unclear. e. miRNA within the mature RISC binds to its target mRNA sites. Adapted from Stroynowska-Czerwinska, Fiszer, and Krzyzosiak 2014.

Argonautes

The RISC-complex may contain any of the four mammalian argonaute proteins, however only AGO2 has endoribonuclease activity and so is the only one that can cleave mRNA. The other argonautes are involved in silencing the mRNA target by translational repression, mRNA deadenylation or one of the other mechanisms briefly described in the “Modes of miRNA action” section.

Post-miRNA processing

These guidelines do not intend to cover the annotation of proteins involved in the miRNA-mediated mechanisms of gene expression regulation after formation of the mature RISC, only the miRNAs, however these proteins are listed in Table 1 as it may be useful to recognise them in context.

Table 1. The proteins involved in miRNA-mediated mechanisms of gene expression regulation. Adapted from Table 1 in Stroynowska-Czerwinska, Fiszer, and Krzyzosiak 2014, which contains the full table including references. No attempt has been made to suggest annotations for these proteins in these guidelines.

Modes of miRNA action

In order to be able to correctly identify GO terms that can be used to describe the functional roles of miRNAs, we must first understand the mechanisms by which miRNAs can regulate gene expression (Figure 4).

These guidelines are primarily for annotating proteins and miRNAs involved in gene silencing via the interaction between a miRNA-containing RISC complex and the 3’UTR of its target mRNA, causing either mRNA cleavage, translational repression (initiation or elongation block) or mRNA deadenylation followed by decapping and degradation of the mRNA. Other mechanisms are briefly described later for information. It should be noted that many of these mechanisms and their dependencies are still not fully understood (Wilczynska and Bushell 2014), therefore a definitive description cannot be given herein. When annotating, however, the curator should always capture the available experimental evidence and consider the intentions and interpretations of the author.

Figure 4. Post-transcriptional modes of miRNA action. miRNAs may promote degradation of mRNA via deadenylation and decapping or mRNA cleavage as well as repress translation of mRNAs by initiation or elongation block.

mRNA cleavage

An mRNA is cleaved when there is near-perfect sequence complementarity between the miRNA and its target mRNA, this is most common in plants. The miRNA guides the RISC complex to the target mRNA, which is cleaved by AGO2. mRNA cleavage can only be mediated by AGO2 as AGO1, 3 and 4 do not have endoribonuclease activity.

Translational repression

Translational repression occurs when there is imperfect sequence complementarity between the miRNA and mRNA target. The RISC complex binds to the 3’UTR of the target mRNA and recruits additional factors, e.g. DDX6 (see Table 1), that interact with translation factors to interfere with translation initiation and/or elongation (Figure 4). Translational repression may lead to deadenylation and degradation.

mRNA deadenylation

This mechanism is one of the most common modes of gene silencing by miRNA (Guo et al. 2010) and occurs when there is imperfect complementarity between the miRNA and mRNA target. Deadenylation is the shortening of the poly(A) tail of the mRNA, and once the tail reaches a minimum length the mRNA is decapped and subsequently degraded. Deadenylation can also lead to translational repression of the mRNA.

Figure 5 depicts those proteins involved in mRNA deadenylation and decapping that follows binding of the miRNA to its target mRNA.

Figure 5. miRNA involvement in mRNA deadenylation and decapping.

Other miRNA-mediated mechanisms of gene regulation

In addition to targeting of the 3’UTR of mRNAs, miRNAs have been shown to regulate gene expression in other ways, as follows.

Alternative translational repression

miRNAs have been shown to repress translation by targeting either the 5’UTR or coding regions of their target mRNAs as opposed to the usual 3’UTR regulation, e.g. miR-20a targets binding sites in PRKG1 to reduce its expression (Zeng et al. 2014). See also Fang and Rajewski 2011, Hausser et al. 2013 and references in Introduction of Zhang et al. 2014.

Translational activation

miRNAs have also been shown to be capable of activating translation by targeting 5’UTR or 3’UTR of mRNAs (see references in Introduction of Zhang et al. 2014), although this has only been shown in a few studies.

Regulation of transcription

miRNAs are also implicated in both the positive and negative regulation of transcription of mRNAs (see references in Introduction of Zhang et al. 2014). See the “Annotating miRNAs that activate gene expression” section for an example of positive regulation of transcription by miRNAs.

miRNA regulation

There is evidence that miRNAs may target other miRNAs to regulate their processing (Cipolla 2014).

As most of these additional mechanisms have not been as well studied as the miRNA role in post-transcriptional gene silencing via the 3’UTR, we will not provide detailed guidelines here on how to annotate each of these mechanisms.

Experimental approaches used to modulate specific miRNA levels

In order to choose appropriate GO terms and evidence codes to describe the effect of miRNAs on their targets, it is important to understand how miRNAs can be modulated. There are a variety of methods for changing the levels of miRNAs (Figure 6) and the most common ones are summarised here. This topic is also reviewed in van Rooij, Marshall, and Olson 2008.

Increasing miRNA levels

1. Pre-miR or miRNA mimics: synthetic RNAs, which mimic a specific miRNA. Often the pre-miR or mimic sequence that is synthesised is identical for several species, in these cases it would be acceptable to make an IDA annotation for any of the species that have this sequence, as appropriate for the experiment being performed. E.g. if the pre-miR is transfected into a mouse, then the annotation may be made to the mouse miRNA, but if the same pre-miR was transfected into a human cell, the annotation would be made to the human miRNA.

 A note on miRNA mimics from Dharmacon: miRNA mimics from the Dharmacon company are widely used in the
 literature. Dharmacon have confirmed that mimics that are based on mature miRNAs between 18-25 
 nucleotides in length have an identical sequence to the miRBase identifer that is cited on the 
 product page for that mimic (see example below). Any mimics that are shorter or longer (>28nt) than
 this may have design constraints on the sequence. It is recommended to contact Dharmacon
 (ts.dharmacon.eu@ge.com) to discuss specific miRNA sequences that fall into the short/long category.
 E.g. The sequence of the miRIDIAN microRNA Mouse mmu-miR-92a-3p mimic
 is 100% identical to the cited miRBase Accession MIMAT0000539.


2. Pri-miRNA: the pri-miR sequence is inserted into the activation system and transfected into whole animals or cells (see example in Castaldi et al. 2014).

a. activated in vivo via an inducible activation system (e.g. TetON http://en.wikipedia.org/wiki/Tetracycline-controlled_transcriptional_activation)

b. activated in vivo using specific promoters to overexpress miRNA in a cell-type or tissue specific manner

Decreasing miRNA levels

Since a given microRNA typically regulates a whole plethora of mRNAs, the consequences of some of these approaches are difficult to predict and may lead to off-target effects. The effects are often transient and a high dose of ‘inhibitor’ is needed for mammalian targets.

miRNA activity may be decreased by targeting the miRNA as in the following;

1. Antisense-miRNAs or antagomirs (e.g. locked nucleic acid; LNA) are chemically modified oligonucleotides that bind specifically to particular mature microRNAs, therefore making them unavailable for binding to their targets (Ebert, Neilson, and Sharp 2007).

2. Targeted deletion/knockout of a single miRNA by homologous recombination: the miRNA is sequence replaced, e.g. with neomycin resistance cassette, and homozygous animals generated (stem cells transfected with construct are injected into blastocysts from which animals are generated) (For examples, see: Zhao et al. 2007 and S. Wang et al. 2008.

3. miRNA eraser refers to the continuous production of the miRNA antisense sequence causing complete knockdown of the targeted microRNA in vivo or in vitro. Endogenous miRNA expression is eliminated/reduced (For example of knockdown see Figure 2 in Sayed et al. 2008; see also: http://techfinder.rutgers.edu/tech?title=Novel_%22miRNA_eraser%22_and_expression_vectors).

miRNA activity may be decreased by targeting the mRNA as in the following;

4. miRNA sponge is a decoy mRNA containing multiple miRNA binding sites that can either bind with all members of a miRNA seed family or bind with a specific miRNA. This leads to functional inhibition of the miRNA family or specific miRNA without affecting their endogenous expression (Ebert, Neilson, and Sharp 2007). If the sponge is designed to target a seed family, i.e. it will inhibit several related miRNAs, it is not possible to annotate such an experiment as it is unclear which miRNA(s) of the family should be annotated. If the sponge is designed to target a single miRNA, then this may be used to annotate the specific miRNA that is inhibited.

5. Blockmirs are steric antisense blockers that bind to specific microRNA binding sites in the target RNAs to prevent microRNA binding to the same site. They are capable of blocking single microRNA:messenger RNA interactions, i.e. capable of preventing a microRNA from regulating one particular messenger RNA, while still allowing the microRNA to regulate all its other messenger RNA targets. Off-target effects are limited. (Example of use see Young et al. 2013).

  Summary:
  •	Antagomirs bind to specific miRNA sequences: can affect many mRNA targets 
  •	miRNA knockout reduces endogenous transcription of miRNA
  •	miRNA erasers are the antisense miRNA resulting in knockdown of the specific miRNA 
  •	miRNA sponges are decoy mRNAs that bind all members of a miRNA seed family: can affect many mRNA targets
  •	Blockmirs bind to target RNAs: affects only the specific mRNA target

miRNA mimics and antagomirs can be used on a high-throughput scale (see Vanhecke and Janitz 2004). Note the naming convention of “pre-miR” for mimics and “anti-miR” for antagomirs, e.g. pre-29b for miRNA-29b mimic, and anti-29b for miRNA-29b antagomir.

Figure 6. Modulation of miRNAs. Regulation of levels of miRNAs can be achieved by several approaches, shown in the shaded rectangular boxes. See text for a description of each of these.

Evidence codes for modulated miRNAs

The evidence code will be based on the type of modulation applied;

a. when adding more of an miRNA (e.g. by pre-miRNA), the miRNA can be annotated using IDA, even though this is an overexpression of the miRNA, because the sequence is unchanged. If the authors mention that the miRNA is being expressed in cells/tissues it wouldn’t normally be expressed in, consider using IMP

b. when inhibiting the action of a miRNA (e.g. by using antagomir or mutating the sequence), the miRNA should be annotated using IMP

Annotation of an entity regulating expression of a specific miRNA and its mRNA targets

Annotation of an entity regulating levels of a specific miRNA

miRNAs are formed by transcription of the pri-miRNA followed by cleavage of pri-miRNA to pre-miRNA then export to the cytoplasm where they are cleaved by DICER1 to mature miRNAs (Figure 1).

The miRNAs are not “biosynthesised”, consequently there is no GO term for miRNA biosynthetic process. The term “miRNA metabolic process” (GO:0010586) should not be used for biogenesis of miRNA as this term only covers miRNA catabolism and the formation of miRNA isoforms (i.e. the term “mature miRNA 3’end processing” refers to creation of isoforms from the mature miRNA). Instead, the terms that should be considered are; for the transcription step, “pri-miRNA transcription from RNA polymerase II promoter” (GO:0061614) and its regulation children and for the maturation events, terms such as “primary miRNA processing” (GO:0031053) and “pre-miRNA processing” (GO:0031054) should be considered (these are child terms of “RNA metabolic process” rather than “miRNA metabolic process” as the mature miRNA is not being processed).

When choosing an appropriate term, consider the type of entity that is causing the altered levels of miRNA; if it is a transcription factor, then it’s likely to be regulating the transcription of the pri-miRNA therefore an appropriate term for regulation of transcription can be used, as mentioned above. Similarly, if the protein added is part of the miRNA processing pathway, e.g. DICER1 or DROSHA etc., then you can use an appropriate term for that part of the pathway, e.g. "primary miRNA processing" for DROSHA (see Figure 1 for other examples).

However, in an experiment that shows altered levels of miRNA following treatment with an entity that is not part of the known miRNA transcription or processing pathways, we don't know which part of miRNA biogenesis is being affected. We must therefore use the “regulation of gene expression” (GO:0010468) terms that cover both transcriptional and posttranscriptional mechanisms.


Example

In PMID:22269326, the authors describe that addition of TGFβ1 to human aortic smooth muscle cells (hASMCs) and human aortic adventitial fibroblasts (hAFBs) causes a decrease in levels of miR-29b as measured using an miRNA assay kit for hsa-miR-29b. The decrease is more prominent in hAFBs (Figure 3a in PMID:22269326).

From this experiment, it is not clear what part of miRNA biogenesis TGFβ1 is affecting. It could be either transcription of pri-miR-29 or posttranscriptional events. We should therefore annotate TGFβ1 to “negative regulation of gene expression” (GO:0010629), which covers both transcription and miRNA processing terms, with IDA evidence. We can indicate both the target of the regulation and the cell type in which this process happens in the annotation extension with the following annotation

  Object:		      UniProtKB:P01137 (human TGFβ1)
  GO term:		      GO:0010629 (negative regulation of gene expression)
  Evidence:		      IDA
  Annotation Extension:       has_input(RNAcentral:URS000024463E_9606 human miR-29b),occurs_in(CL:0002547 fibroblast of the aortic adventitia)

Annotating an entity regulating levels of an mRNA, when mediated by miRNA

In the experiment shown in Figure 3a in PMID:22269326 the role of TGFβ1 was not clear, beyond affecting the levels of miR-29b. In a subsequent experiment, Maegdefessel et al. went on to measure the levels of the predicted target mRNAs of miR-29b, using a combination of TGFβ1 treatment and either pre-29b or anti-29b to increase or decrease levels of miR-29b (Figure 3b in PMID:22269326).

Ordinarily, an effect on mRNA levels following treatment with a protein would be annotated with “regulation of transcription” GO terms. However, in Figure 3b of PMID:22269326 it is clear that miRNAs are also involved in the overall mRNA level of COL1A1 and COL3A1. As miRNAs have been shown to regulate gene expression at the level of both transcription and post-transcription we have to consider, given the available evidence, whether the action of TGFβ1 on mRNA levels (the addition of TGFβ1 increases the levels of mRNA) is due to regulation of transcriptional or post-transcriptional mechanisms.

COL1A1 and COL3A1 mRNAs have validated miR-29b binding sites in their 3’UTR, therefore we can assume that this miRNA is likely to be silencing these genes through a post-transcriptional mechanism. Consequently, we can choose the term “negative regulation of gene silencing by miRNA” (GO:0060965) for TGFβ1 and include the transcript identifiers for the collagens in the annotation extension, since Figure 3b of PMID:22269326 demonstrates that miR-29b is also regulating mRNA levels of these collagens, which are validated targets of miR-29b (see below for annotations associated with the miRNA).

If we combine the evidence from Figure 3a and 3b from PMID:22269326 we can also annotate TGFβ1 to “negative regulation of production of miRNAs involved in gene silencing by miRNA” (GO:1903799). Although these two terms are related in the ontology, we are making different statements in the annotation extension (see below) and so these are unique annotations. Ideally, we should be able to combine these statements, but currently we have no mechanism for including the mRNA in the second annotation as it would be unclear what the regulation is of.

The annotations would look like this

Annotation 1:

 Object:		 UniProtKB:P01137 (human TGFβ1)
 GO term:		 GO:0060965 (negative regulation of miRNA-mediated gene silencing)
 Evidence:		 IGI
 With/From:		 RNAcentral:URS000024463E_9606 (human miR-29b)
 Annotation Extension: has_input(UniProt:P02452 human:COL1A1 gene)|has_input(UniProt:P02461 human:COL3A1 gene)

When using the has_input relations with a miRNA-regulated gene, the UniProt or Model organism gene ID should be used in the annotation extension.

Annotation 2:

 Object: 	        UniProtKB:P01137 (human TGFβ1)
 GO term:	        GO:1903799 (negative regulation of miRNA processing)
 Evidence:	        IGI
 With/From:		RNAcentral:URS000024463E_9606 (human miR-29b)
 Annotation Extension: has_input(RNAcentral:URS000024463E_9606 human miR-29b)

When looking at gene silencing by miRNAs, it is important to note that this is negative regulation of gene expression. Gene expression covers both transcriptional and posttranscriptional events, whereas gene silencing by miRNA only covers post-transcriptional events, therefore the regulation of gene expression term (or negative/positive terms) should be chosen when annotating an entity when the miRNA role in regulating an mRNA is not clear.

A note on controls

Ideally authors will show all of their controls, but in some figures controls are missing. In Figure 3b from PMID:22269326 the controls for TGFβ1 alone or scr-miR alone are not shown, which might make it difficult to judge the true effect of anti-29b and pre-29b. However, we should take into account the authors interpretation of the experiment in our decision of whether or not to annotate such experiments.

Annotating a miRNA regulating levels of an mRNA

The GO terms that should be used for the action of the miRNA on gene expression, depending on available evidence:

Biological Process terms

  • “miRNA-mediated post-transcriptional gene silencing” (GO:0035195)

OR

  • “negative regulation of gene expression” (GO:0010629)

Molecular Function terms

  • “mRNA base-pairing translational repressor activity” (GO:1903231).
  • The term "mRNA 3'UTR binding" GO:0003730 is not required as almost all miRs act by binding this region of the target mRNA. However, if the miR binds a different region of the mRNA then create the relevant specific mRNA binding term

The validated target mRNA may also be included in the annotation extension field of both Biological Process and Molecular Function if there is appropriate experimental support. The GO Consortium has agreed to use UniProt IDs or the Model Organism Gene IDs to represent the target mRNAs

Predicted vs. validated miRNA targets

For GO annotation purposes, we are classifying targets as either predicted or validated. These are explained further below.

Predicted target

A predicted target is one where a sequence alignment of the mRNA and miRNA shows the predicted binding site of the miRNA, this may be found in a paper or in one of the miRNA target prediction databases (e.g. TargetScan, MiRBase, DIANA-microT, GOmir, MiRWalk, etc.). If the only evidence of a target is prediction then this is not captured as a GO annotation.

Validated targets

Validated targets are those that have undergone some kind of experimental assay to demonstrate that the miRNA has an effect on the target mRNA. An increasing number of experimental methods are available for testing either the binding or the regulation aspects of miRNA targets, some of which are classified as high-throughput (HTP) (Chou et al., 2016 and Thomson et al., 2011). To assist biocurators in deciding whether the evidence for a miRNA:mRNA functional interaction (i.e. binding and regulation) is sufficient to create an annotation, a list of commonly reported methodologies is provided below, including a description of the method and whether it is sufficient on its own to demonstrate the binding and regulation aspects or whether additional experimental evidence is required. We recommend that in cases which combine two methods for demonstrating functional interaction, only one of these should be a HTP method as indicated on the list; as technologies improve this decision may be re-visited. A HTP method by itself should not be used to curate a functional interaction. For example:

  • a reporter assay alone is sufficient to demonstrate binding and regulation
  • a CLASH experiment (HTP) can only demonstrate binding of the miRNA:mRNA, therefore additional evidence such as a qRT-PCR (non-HTP) demonstrating the regulation of the mRNA levels must also be required.

Several of the methods below only measure one of these aspects and often combine the method with a complimentary method to suggest functional targets. We suggest that in the cases which combine two methods for validation of miRNA targets, only one of these can be a high-throughput method as indicated on the list, as technologies improve this decision may be re-visited. Those that are marked as do not annotate should not be used at all as evidence even in combination with a method showing the complimentary binding or mRNA/protein levels. E.g. if microarray was used in combination with CLASH, this wouldn’t be sufficient evidence to annotate as a validated miRNA:mRNA target.

The list comprises the most common methods encountered in the literature, any method not listed should be considered for discussion with the BHF-UCL miRNA annotation team. Reviews of experimental technologies used to validated miRNA:mRNA functional interactions; Thomson et al., 2011 (includes caveats of over-expressing miRNAs) and Kuhn et al., 2008.

For the purposes of GO annotation, we have defined two types of validated target, “binding target” and “other target”, the evidence required for each is explained below. Both types of validated target may or may not have been predicted.

Validated binding target
  • In order to create a target binding annotation it is necesssary to have experimental evidence which demonstrates (validates) the binding.
  • In order to create a gene silencing by microRNA annotation it is necesssary to have experimental evidence which demonstrates (validates) the binding and that the expression of the gene is regulated by the binding.

A validated binding target is one that has been shown to be directly acted upon by the miRNA, i.e. experiments have been performed to demonstrate binding between the miRNA and mRNA and regulation of the levels of mRNA. The following experiments demonstrate binding;

1. Reporter assay: (see example in Nicolas 2011) the 3’UTR of the mRNA target is fused to a luciferase reporter gene, this is transfected into cells together with the miRNA, the expected result for a silencing miRNA is a decrease in luciferase activity due to repression of the mRNA by the miRNA. Ideally, this will be confirmed by mutating the miRNA seed sequence site in the 3’UTR of the mRNA to check that the sequence integrity is essential to regulate levels of the luciferase. A reporter assay is deemed to be good enough evidence that the miRNA binds the mRNA as it is unlikely that this effect of the miRNA on the reporter gene is indirect (pers. comm. Anastasia Kalea, UCL) (Recommendation: this may be used as evidence for both binding and regulation with IDA evidence).

2. CRISPR/CAS9 (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR-associated protein-9 nuclease) with subsequent measurement of protein levels in vivo: (see example in Bassett et al. 2014) Mutation (deletion) of miRNA response elements in the targeted gene disrupts binding between miRNA and mRNA, subsequent measurement of protein levels of the target indicates regulation of the target. The mutations are generated in vivo, therefore if the silencing is cell-specific, this can easily be seen, e.g. the target may be silenced in one cell type but not another. (Recommendation: this may be used as evidence for both binding and regulation, with IMP evidence).

3. Affinity purification: (see examples in Vo et al. 2010; Orom and Lund, 2007) Either the miRNA or mRNA is tagged, e.g. with biotin or GFP, and then an affinity purification method is used to identify the binding mRNA or miRNA. This type of method demonstrates direct binding between the miRNA and mRNA. (Recommendation: only annotate these if accompanied by further experiments demonstrating that the levels of mRNA are altered in response to the miRNA, as this method only demonstrates binding).

High-throughput binding assays;

4. CLASH (Cross-linking ligation and sequencing of hybrids): (see example in Helwak and Tollervey, 2014): incorporates a ligation step, concatenating the miRNA to the mRNA binding region. The derived chimeric miRNA:mRNA fragments are subsequently sequenced and bioinformatically separated for the concurrent identification of the targeted mRNAs, binding sites and interacting miRNAs. The class of CLIP-Seq/CLASH experiments can reveal thousands of miRNA:gene interactions in each analysis and has significantly altered the scope and scale of relevant research projects. (Recommendation: only annotate these if accompanied by further experiments, which are NOT high-throughput, demonstrating that the levels of mRNA are altered in response to the miRNA, as this method only demonstrates binding).

5. AGO-iPAR-CLIP (Argonaute-in vivo Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation): (see example in Grosswendt et al., 2014): similar to CLASH above. (Recommendation: only annotate these if accompanied by further experiments, which are NOT high-throughput, demonstrating that the levels of mRNA are altered in response to the miRNA, as this method only demonstrates binding).

It must be noted, however, that these methods only provide an indication that the miRNA is able to bind the mRNA not that it will bind in a physiological context, where there may not be coincident expression of the miRNA and mRNA and also where the concentrations and stoichiometry of the miRNA and mRNA may be different.

Validated other target

A validated other target is one that has not conclusively been shown to be directly acted upon by the miRNA, i.e. it is not known whether the miRNA binds to the mRNA, but there is experimental evidence that the mRNA levels are affected by the miRNA using one or more of the techniques in the list below.

1. Western blot: after addition or removal of the miRNA to the assay system, a Western blot is performed to measure the levels of the target protein. (Recommendation: only annotate these if accompanied by evidence that there is binding between a given miRNA and mRNA, as this method only demonstrates changes in levels of proteins in response to a miRNA).

2. qRT-PCR: after addition or removal of the miRNA to the assay system, a qRT-PCR is performed to measure the levels of the target mRNA. (Recommendation: only annotate these if accompanied by evidence that there is binding between a given miRNA and mRNA, as this method only demonstrates changes in levels of mRNAs in response to a miRNA).

High-throughput regulation assays;

3. pSILAC (Pulsed stable isotope labeling by amino acids in cell culture): (see examples in Boettger et al. 2009; Selbach et al., 2008) - high-yield generalization of ELISA assays and western blots. pSILAC quantifies differences in protein production between two samples. Can correlate reduced protein levels in response to a particular miRNA with whether the target has 3’UTR seed site. Does not measure binding directly (equivalent to doing a western blot - you can see the effect on the protein levels, but you can’t see if this is direct or not). (Recommendation: only annotate these if accompanied by evidence that there is binding between a given miRNA and mRNA by a non-high-throughput method, as this method only demonstrates changes in levels of proteins in response to a miRNA).

4. Degradome Sequencing: (see example in German et al., 2008) Identification of mRNA cleavage products by the parallel analysis of RNA ends (PARE). Since mRNA cleavage is uncommon in mammals this method is primarily used in plants. This method shows binding and regulation (i.e the end point of the interaction, which is cleavage), but since it is a high-throughput method it is recommended that further, non-high-throughput, evidence of the miRNA:mRNA functional interaction is found.

Methodologies that should not be used as evidence of miRNA:mRNA functional interaction

1. Microarray: after addition or removal of the miRNA to the assay system, a microarray is performed to detect differentially expressed mRNAs - a high-throughput version of qPCR and northern blotting

2. NGS (Next Generation Sequencing): provides sequencing of miRNAs, high-throughput/increased accuracy. No indication of target or effect on mRNA levels. Usually used in combination with a binding experiment, e.g. PAR-CLIP/HITS-CLIP (see below), to sequence the miRNAs/mRNAs bound. (Recommendation: do not annotate).

3. Small RNA-seq/RNA-seq: Types of NGS, sequence and expression levels of miRNAs measured, provides no information on target or its levels. (e.g. Landgraf et al., 2007 and Chou et al., 2015) (Recommendation: do not annotate)

4. HITS-CLIP (High-throughput sequencing of RNA isolated by crosslinking immunoprecipitation): transcriptome-wide map of miRNA binding sites, however, the identified regions are usually wide and perplex the identification of the exact miRNA binding location, which is performed algorithmically (e.g. Chi et al., 2009).(Recommendation: do not annotate)

5. PAR-CLIP (Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation): A modified CLIP-Seq methodology demonstrating binding. Compared to the results obtained by HITS-CLIP, the boundaries of the identified binding locations are sharper and significantly narrower. Despite the accurate detection of the crosslinked region, these methods cannot directly reveal the specific miRNA participating in the interaction, which has to be bioinformatically identified (e.g. Hafner et al., 2010). (Recommendation: do not annotate)

Note that, although the intentions of the author should be considered, it is recommended that the curator does not rely on an author statement that an mRNA is a target of a miRNA. There should be evidence that the target is predicted or validated, e.g. a sequence alignment or experiment.

Figure 7 will help you decide which GO term should be used and what format the annotation extension should take, depending on the evidence you have.

Decision tree to assist with curation of miRNA targets

Link to powerpoint MiRNA Decision Tree 2023 document File:MiRNA Decision Tree 2023.pptx

Link to MiRNA Decision Tree 2023 pdf File:MiRNA Decision Tree.pdf

Figure 7. Updated Decision Tree for the GO terms and annotation extensions used for capturing targets of miRNAs. The types of evidence in the blue boxes are described further in the text. Information about the cell or tissue that the interaction occurs in can be captured for example in the annotation extension field using: occurs_in (insert cell ontology ID and/or UBERON ID based on cell and/or tissue type respectively). This information can be included when inhibition of the miRNA leads to an increase in the endogenous mRNA/protein level. Slightly weaker evidence to support inclusion of the cell or tissue information would be evidence that the miRNA is expressed in these cells/this tissue AND that the authors are confident that the miRNA target is also being regulated in these cells/this tissue. For example, when the addition of the miRNA leads to a decrease in the endogenous mRNA/protein level. (This figure is reproduced from Antonazzo et al. article in preparation, July 2023).

Note: to be able to indicate any target in a GO annotation, the prediction evidence does not need to be in the paper you are annotating, it could be from another paper or a prediction database. When looking at predictions or validations in a prediction database, it is recommended to check the references cited for the target as they may serve as a good source of experimental annotation. Additionally, it has been found that some of the cited papers do not support the validation of the targets. Annotations to “mRNA base-pairing translational repressor activity” (GO:1903231) should ONLY be made from the paper containing the experimental evidence, however it is acceptable to indicate in the annotation extension of a “gene silencing by miRNA” annotation that the target of gene silencing is direct (by using has_input) if the evidence for binding is in another paper.

Example

The authors of PMID:24807785 (Castaldi et al. 2014) investigate whether regulators of the β-adrenergic receptor (β1AR) pathway are direct targets of miR-133. In Figure 1 they show alignments of miR-133 with the target sequences as well as performing luciferase reporter assays to demonstrate the 3’UTRs of the target genes are inhibited by miR-133. They also demonstrate, using immunoprecipitation, that miR-133 could bind the three targets (Figure 1D of PMID:24807785).

Since the authors have shown sequence alignments for miR-133 and its targets and have demonstrated the binding of miR-133 to the 3’UTRs of the specified mRNAs as well as downregulation using luciferase reporter, we can class these as validated binding and predicted targets of miR-133.

We can therefore make the following annotations

Annotation 1:

 Object:		RNAcentral:URS00004C9052_10090 (mouse miR-133a)
 GO term:		GO:0035195 (miRNA-mediated gene silencing)
 Evidence:		IDA
 Annotation Extension: has_input(UniProt:P34971 mouse:adrb1 gene)|has_input(UniProt:Q01341 mouse:adcy6 gene) |has_input(UniProt:P68181 mouse:|prkacb gene)

Annotation 2:

 Object:		RNAcentral:URS00004C9052_10090 (mouse miR-133a)
 GO term:		GO:1903231 (mRNA base-pairing translational repressor activity)
 Evidence:		IDA
 Annotation Extension: has_input(UniProt:P34971 mouse:adrb1 gene), part_of(GO:0035195 miRNA-mediated gene silencing) |has_input(UniProt:Q01341 mouse:adcy6 gene), part_of(GO:0035195 miRNA-mediated gene silencing)|has_input(UniProt:P68181 mouse:prkacb gene), part_of(GO:0035195 miRNA-mediated gene silencing)

Example

In Figure 3b of PMID:22269326 we saw the effect of TGFβ1 and miR-29b on the mRNA levels of collagen genes. The authors have indicated that the collagen genes are predicted targets of miR-29b and they have shown that the levels of the collagen mRNAs change when miR-29b is modulated. COL1A1 and COL3A1 are therefore validated other and predicted targets of miR-29b, so from this evidence we can make this annotation

 Object:		RNAcentral:URS000024463E_9606 (human miR-29b)
 GO term:		GO:0035195 (miRNA-mediated gene silencing)
 Evidence:		IGI
 With/From:		UniProtKB:P01137 (human TGFβ1)
 Annotation Extension: has_input(UniProt:P02452 human:COL1A1 gene)|has_input(UniProt:P02461 human:COL3A1 gene)

However, COL1A1 and COL3A1 have previously been shown to be validated binding targets of miR-29b by luciferase assay in (Steele, Mott, and Ray 2010) (as indicated in miRTarBase: http://mirtarbase.mbc.nctu.edu.tw/), therefore we could instead use the annotation extension relation has_input in the above annotation.

Note: If the miRNA has been shown to target binding sites other than those in the 3'UTR this may still be given the "gene silencing by miRNA" GO term if there is evidence of regulation of gene expression and identification of predicted binding sites for the miRNA, e.g. addition of the miRNA causes downregulation of the mRNA but a luciferase reporter assay shows no regulation, however binding sites in the coding region are identified (see Zeng et al. 2014 for an example of this).

Using the more specific child terms of gene silencing by miRNA

In some circumstances, when the authors have shown the exact mechanism of silencing, i.e. translational repression, deadenylation or mRNA cleavage, it is possible to use the child terms of the Biological Process term “gene silencing by miRNA”. Translational repression and mRNA deadenylation are the most common mechanisms for animal miRNAs, whereas mRNA cleavage is the major mechanism for plant miRNAs (Axtell, Westholm, and Lai 2011). The following three sections give examples for each of these scenarios. As with the previous examples, if there are additional experiments demonstrating binding of the miRNA to the mRNA then the Molecular Function term “mRNA binding involved in posttranscriptional gene silencing by miRNA” (GO:1903231) may also be used.

Translational repression

Translational repression occurs when there is imperfect sequence complementarity between the miRNA and mRNA target.

Example

The authors of PMID:16380711 (Chen et al. 2006) used a luciferase assay to confirm the interaction between miR-1 and miR-133 with HDAC4 and SRF mRNAs, respectively. However, they observed that although overexpression of miR-1 or miR-133 led to downregulation of HDAC4 and SRF proteins, respectively, the mRNA levels were not affected. Therefore it was concluded that the miRNAs action led to inhibition of mRNA translation (see Figure 4 in PMID:16380711)

The more specific biological process GO term can be applied based on this evidence:

Annotation 1:

 Object:			RNAcentral:URS00001DC04F_10090 (mouse miR-1a)
 GO term:	        GO:0035278 (miRNA-mediated gene silencing by inhibition of translation)
 Evidence:		IMP
 Annotation Extension:	has_input (UniProt:Q6NZM9 mouse HDAC4 gene)

Annotation 2:

 Object:			RNAcentral:URS00004883E0_10090 (mouse miR-133)
 GO term:	        GO:0035278 (miRNA-mediated gene silencing by inhibition of translation)
 Evidence:		IMP
 Annotation Extension:	has_input (UniProt:Q9JM73 mouse SRF gene)

mRNA deadenylation

This mechanism is one of the most common modes of gene silencing by miRNA (Guo et al. 2010) and occurs when there is imperfect complementarity between the miRNA and mRNA target. However, there are very few articles published that describe this specific mechanism and therefore the GO term GO:0035279: miRNA-mediated gene silencing by mRNA destabilization should be applied.

Example

The authors of PMID:17671087 (Wakiyama et al. 2007) investigated Let-7 miRNA-mediated mRNA deadenylation by analysing RNaseH cleavage of the poly(A) tail. After treatment, in the presence of Let-7, the bands of mRNAs shifted to almost the same position as non-adenylated RNAs, indicating deadenylation had taken place (see Figure 4 in PMID:17671087).

The annotation that can be made from this evidence is

Annotation 1:

 Object:			RNAcentral:URS0000416056_9606 (human let-7)	
 GO term:	        GO:0035279 (miRNA-mediated gene silencing by mRNA destabilization)
 Evidence:		IMP

Note: No annotation extension is given in this example as the mRNA used was from Drosophila, if the mRNA used was from the same species as the miRNA, then the extension would use the has_input relation with the UniProt or Model Organism gene identifier of the target mRNA.

mRNA cleavage

Endonucleolytic cleavage of target mRNAs is more widely found in plants, although there are some examples of this occurring in animals (Axtell, Westholm, and Lai 2011). The sequence complementarity between miRNA and target should be perfect or near-perfect for cleavage to occur.

Example

In PMID:25794935, Arabidopsis miR-847 was shown to have a complementary sequence in the mRNA encoding IAA28 (J.-J. Wang and Guo 2015). The authors tested the ability of miR-847 to cleave IAA28 mRNA and a mutant version of IAA28 mRNA. In Figure 2 of PMID:25794935, the results show cleavage of IAA28 mRNA in the presence of miR-847 but no cleavage of the mutant version.

The annotation that can be made from this evidence is

Annotation 1:

 Object:			RNAcentral:URS00001CE296_3702 (Arabidopsis miR847)
 GO term:	        GO:0035279 (mRNA cleavage involved in gene silencing by miRNA)
 Evidence:		IDA
 Annotation Extension:	has_input(TAIR:locus:2180557 Arabidopsis IAA28 gene)

Annotating MicroRNA clusters

MicroRNA (miRNA) clusters are defined as miRNA genes, which are located on a chromosome at the maximum inter-miRNA distance (MID) of 10 kb (miRBase). It is believed that the clustered miRNA genes are typically transcribed as one messenger RNA (mRNA) molecule from one polycistron (polygene) under the control of one promoter. This has indeed been demonstrated for the mouse miR-17-92 cluster (Woods et al., 2007). MiRNAs within a cluster are often homologous and they act in concert to coordinate regulatory functions.

Several studies demonstrated that the number of miRNA clusters identified within the human genome ranges from 31 (at MID of 3 kb) to 98 (at MID of 10 kb) (Yu et al., 2006, Altuvia et al., 2005 and Sun et al. 2013). For comparison 61 miRNA gene clusters were identified in the mouse genome at MID of 10 kb (Sun et al. 2013).

The identified clusters were located either at intergenic regions of the genome, or, in about a third of cases, they were present within the introns or 3’ UTR regions of protein-coding genes (Altuvia et al., 2005). The intragenic miRNA genes may be expressed under the control of the promoter of the host mRNA, from which they are subsequently excised as intronic lariats (Cullen 2004). More than two thirds (Yu et al., 2006) of miRNA gene clusters contained homologous genes (miRNA homo-clusters; Sun et al. 2013), and about a half formed paralogous cluster groups with other miRNA clusters (Yu et al., 2006).

For further details on evolution and conservation as well as on how the evolutionary outcomes are reflected within individual modern genomes please refer to the following sources: Zhang et al., 2009 and Sun et al. 2013.

The great conservation of miRNA clusters within and between species emphasises the functional importance of these genomic regions. It has been suggested that miRNAs within a cluster cooperate to regulate gene expression, which is more efficient than regulation by discrete miRNAs (Sun et al. 2013). Not surprisingly, the more conserved clusters are more likely to be associated with diseases. Furthermore, there is evidence to suggest that many more miRNAs than previously believed probably exist in clusters (Altuvia et al., 2005).

MiRNA clustering has naturally also had an impact on experimental design and, consequently, annotation of miRNAs and their targets. The homologous nature of clustered miRNAs may sometimes prompt scientists to evaluate them as one entity.

For instance, Figure 2 in publication PMID:19390056 (Brock et al., 2009) presents the results of luciferase reporter assays, in which the effect of the whole miR-17-92 cluster on target gene expression is evaluated (Fig. 2A) prior to testing the activities of individual miRNAs from this cluster (Fig. 2B). Since outcomes of discrete miRNA-mRNA binding are presented in Figure 2B it is preferable for biocurators to use identifiers for the specific miRNAs, available from the RNAcentral database, for annotation, particularly in this case because only two of the four mature miRNAs affected gene silencing of the target e.g.;

Annotation 1:

Object:                RNAcentral:URS00002075FA_9606 (human miRNA-17-5p)  
GO term:               GO:0035195 (gene silencing by miRNA)
Evidence:              IDA  
Annotation Extension:  has_input(UniProt:Q13873 human BMPR2)

Annotation 2:

Object:                RNAcentral:URS00002075FA_9606 (human miRNA-17-5p)  
GO term:               GO:1903231 (mRNA binding involved in posttranscriptional gene silencing)
Evidence:              IDA  
Annotation Extension:  has_input(UniProt:Q13873 human BMPR2)

However, had the researchers studied only the effects of the whole miR-17-92 cluster (Figure 2A), then biocurators would still have to use identifiers that describe the gene product(s) which has/have activity within the whole miRNA cluster, i.e. to use the RNAcentral identifiers for discrete mature miRNAs. If the authors describe which of the possible mature miRs are predicted to bind the target then curate only that/those miR(s). If more than 1 miR in the cluster is predicted to have the activity but the experiments only use the whole cluster then use the IGI evidence code and include the other 'predicted' miRs in the WITH field.

For example, in PMID: 18723672 Figure 3 COS cells were transiently transfected with luciferase reporters linked to the 3′UTR sequences of the several different genes. Increasing amounts of the miR-29b-1/miR-29a cluster repress the expression of luciferase. The annotations are as follows.

Annotation 1:

Object:               RNAcentral:URS000024463E_10090 (mouse miR-29b-3p)
GO term:              GO:0035195 (gene silencing by miRNA)
Evidence:             IGI
WITH:                 RNAcentral:URS00002F4D78_10090  (mouse miR-29a-3p)
Annotation Extension: has_input(UniProt:P54320 Eln)

Annotation 2:

Object:               RNAcentral:URS00002F4D78_10090 (mouse miR-29a-3p)
GO term:              GO:0035195 (gene silencing by miRNA)
Evidence:             IGI
WITH:                 RNAcentral:URS000024463E_10090 (mouse miR-29b-3p)
Annotation Extension: has_input(UniProt:P54320 Eln)

Annotation 3:

Object:               RNAcentral:URS000024463E_10090 (mouse miR-29b-3p)
GO term:              GO:0003730 (mRNA 3’UTR binding)
Evidence:             IGI
WITH:                 RNAcentral:URS00002F4D78_10090  (mouse miR-29a-3p)
Annotation Extension: has_input(UniProt:P54320 Eln), part_of (GO:0035195 gene silencing by miRNA)

Annotation 4:

Object:               RNAcentral:URS00002F4D78_10090 (mouse miR-29a-3p)
GO term:              GO:0003730 (mRNA 3’UTR binding)
Evidence:             IGI
WITH:                 RNAcentral:URS000024463E_10090 (mouse miR-29b-3p)
Annotation Extension: has_input(UniProt:P54320 Eln), part_of (GO:0035195 gene silencing by miRNA)

Annotating miRNAs that activate gene expression

The role of miRNAs in gene silencing by guiding the RISC to 3’UTRs of target mRNAs is well studied. However, there is also evidence that miRNAs may mediate gene regulation in others ways (see section “Modes of miRNA action”).

Example

In Zhang et al. (2014), Let-7i was computationally predicted to bind the core promoter region of interleukin-2 (IL2) (see Figure 2 in PMID:25336585). The authors then went on to demonstrate that overexpression of Let-7i in HEK293T cells enhanced the IL2 promoter activity (promoter fused to luciferase reporter). They also show an increase in IL2 mRNA and protein levels upon treatment with Let-7i in PLKO and CD4+ T-cells (Fig. 2D,E,G in PMID:25336585) and also a decrease in IL2 mRNA levels upon treatment with a Let-7i inhibitor (Fig. 2H in PMID:25336585).

Mutation of either the Let-7i binding site in the IL2 promoter or of the seed sequence of Let-7i reduced the luciferase activity (Figure 3 in PMID:25336585), demonstrating that the binding of Let-7i to the TATA box in the core promoter of IL2 is sequence-specific.

Using this evidence, we can make the following annotations

Annotation 1:

 Object:		RNAcentral:URS00004023EA_9606 (human Let-7i)
 GO term:	        GO:0000979 (RNA polymerase II core promoter sequence-specific DNA binding)
 Evidence:		IDA
 Annotation Extension:	has_input(UniProt:P60568 human IL2 gene)

Annotation 2:

 Object:		RNAcentral:URS00004023EA_9606 (human Let-7i)
 GO term:	        GO:0045944 (positive regulation of transcription from RNA polymerase II promoter)
 Evidence:		IDA
 Annotation Extension:	has_input(UniProt:P60568 human IL2 gene)

This paper shows quite an in-depth study of the action of the miRNA. If, however, the only experiment shown is an increase in levels of an mRNA or increased transcription of a gene following miRNA treatment, with no indication of binding to the 3’UTR or gene promoter, the GO term used should be “positive regulation of gene expression” (GO:0010628) indicating the UniProt or Model Organism identifier in the annotation extension field using has_input.

Capturing cell and tissue type information

When annotating an experiment where miRNA is over-expressed, e.g. pre-miR, if the cell or tissue it is being over-expressed in normally expresses this miRNA then the cell/tissue information can be added to the annotation extension with the relation occurs_in. If there is no evidence that the cell or tissue normally expresses this miRNA, then do not add this information to the annotation extension.

Example

PMID:25201911 (Zampetaki et al. 2014) describes the transfection of miRNA mimics into smooth muscle cells. Overexpression of miR-195 as well as miR-29b led to a similar reduction in elastin (ELN) mRNA levels (see Online Figure II in PMID:25201911). Additionally, the authors state that the miRWALK algorithm predicts putative binding sites for miR-195 and miR-29b in the coding region and the 3′ untranslated region of ELN, respectively. Therefore, according to our criteria for targets (see “Predicted vs. validated miRNA targets” section), ELN is a validated other and predicted target of both miRNAs.

The authors use mimic miRNAs to increase levels of miR29b and miR-195 in a cell type that is expected to express these miRNAs, smooth muscle cells, therefore the annotation we can make from this evidence for both miRNAs is

 Object:		RNAcentral:URS000024463E_10090 (mouse miR-29b)
 GO term:	        GO:0035195 (gene silencing by miRNA)
 Evidence:		IDA
 Annotation Extension:  has_input(UniProt:P54320 mouse ELN gene), occurs_in(CL:0000192) (smooth muscle cell)

Capturing the downstream effects of the miRNA

When a miRNA silences an mRNA, it causes a decrease in the amount of the protein encoded by that mRNA, and thus often the activity of the encoded protein, therefore we can annotate the miRNA using a term describing the regulation of that activity.

Example

The authors of the same paper that gave us the mRNA targets of miR-133 (Castaldi et al. 2014) went on to investigate the silencing effect of miR-133 on adenylate cyclase activity. They overexpressed miR-133 and found that this decreased the amount and rate of cAMP accumulation (see Figure 3 in PMID:24807785)

Since the authors have also demonstrated that adenylate cyclase inhibits the adrenergic receptor signaling pathway, the annotations we can make from this evidence is

Annotation 1:

 Object:		RNAcentral:URS00004C9052_10090 (mouse miR-133a)
 GO term:	GO:0071877 (regulation of adenylate cyclase-inhibiting adrenergic receptor signaling pathway)
 Evidence:		IDA
 Annotation Extension:	occurs_in(CL:0000746) (cardiac muscle cell)


Note: there is no entry for ADCY6 in the annotation extension for these particular annotations since the authors measured total adenylate cyclase activity, not specifically ADCY6. However, since they measure the activity in cardiomyocytes and miR-133 is expressed in this cell type (shown elsewhere in the paper by the use of an Ad-decoy133 construct to knockdown miR-133 in cardiomyocytes), we can enter the cell type information in the annotation extension.

It is important to note that if the authors state that the miRNA is being expressed in cells/tissues where it would not normally be expressed, the experiment is not physiologically relevant information - miRNAs are known to have different targets in different cell and tissue types (Zhu et al., 2011) - so it would not be appropriate to create an annotation.

If these processes are being studied in the context of a larger process, then it is acceptable to use or create terms that describe that involvement.

To assist the curator in finding appropriate GO terms to use, the curator should also consider the intentions and interpretations of the author, for example if the author shows the effect of a miRNA on a specific process - is this an expected result, does it fit with the validated targets of the miRNA?

Capturing the effect of miRNAs on unknown targets

If it is not known what target(s) the miRNA has, i.e. the authors add a miRNA to a system/animal to see what effect it has, we can annotate to regulation of the process(es) that it affects, even if we have no other evidence for the role of that miRNA.

Example

In PMID:25707426 (Dang et al. 2015), the authors investigate the effect of selected miRNAs on CD16- monocyte cell motility. miR-19a knockdown slowed the movement of these cells compared with control cells (see Figure 4 in PMID:25707426), suggesting miR-19a may regulate genes involved in CD16- monocyte motility.

From this evidence we could make the following annotation even though we do not know the specific targets of miR-19a

 Object:		RNAcentral: URS00001754CF_9606 (human miR-19a)
 GO term:	        GO:2000145 (regulation of cell motility)
 Evidence:		IDA
 Annotation Extension:	results_in_movement_of(CL:0000576) (monocyte)

Plant miRNA biogenesis and action

The canonical plant miRNA biogenesis pathway (Figure 8) shares some commonalities with the mammalian pathway, however there are also some differences (Axtell, Westholm, and Lai 2011).

For instance, in a recent paper (Lauressergues et al. 2015) the authors describe the discovery of microRNA regulatory peptides (miPEPs) that are produced from primary miRNA transcripts. These peptides activate the transcription of the primary miRNAs they derive from, resulting in accumulation of the corresponding mature miRNAs. It is too early to say whether miPEPs are conserved in mammals.

The major difference, however, is that in plants miRNA biogenesis is largely completed within the nucleus. All of the processing steps involving the pri- and pre-miRNAs and carried out by DCL1 occur in the nucleus, whereas in mammals these processes are separated between the nucleus and cytoplasm (see Figure 1). Another difference is that the plant miRNA duplex is stabilized after cleavage by 2’-O-methylation by the HUA ENHANCER 1 protein.

There are also differences in target recognition between mammalian and plant miRNAs (Rogers and Chen 2013; Xie, Zhang, and Yu 2015). Plant miRNAs require perfect, or near-perfect complementarity with the target mRNA, whereas mammalian miRNAs generally only have complementarity with their targets within the seed region. The target mRNAs are silenced most commonly by cleavage, although there are many mRNAs that instead undergo translational repression. Currently, there is no evidence that plant miRNAs are involved in mRNA destabilization by deadenylation.

Figure 8. The canonical plant miRNA processing pathway. The proteins involved in the pathway of miRNA formation are shown together with the GO terms that are expected to be associated with them. Protein names: Not2b: Negative on TATA-less 2B; CDC5: Cell division cycle 5; Mediator complex proteins include: MEDIATOR21 (AT4G04780), MEDIATOR14 (AT3G04740), MEDIATOR25 (AT1G25540), MEDIATOR20a (At2g28230), MEDIATOR20b (At4g09070), MEDIATOR20c (At2g28020); miPEP: miRNA encoded peptide; DCL1: Dicer-like 1; SE: Serrate; HYL1: Hyponastic leaves 1; DDL: Dawdle; TGH: Tough; CBP20: Cap binding complex protein 20; CBP80: Cap binding complex protein 80; HEN1: HUA ENHANCER 1; HST: Hasty; HSP90: Heat shock protein 90; SQN: Squint; AGO: Argonaute. (From Huntley et al. RNA. 2016 May;22(5):667-76).

Curation issues

Potential problems that may be encountered by a curator include the following;

Finding evidence that an mRNA is a predicted target of a miRNA

If an experiment shows that a miRNA binds an mRNA target and has an effect on its levels then it is straightforward to designate the target as validated binding (see Figure 7). However, finding evidence that an mRNA is a predicted target (in order to designate it as a validated other target) can be more difficult. Ideally, there would be a published alignment of the miRNA:mRNA, but in the absence of this it is necessary to look in the miRNA target prediction databases, such as TargetScan, miRBase etc. This is not necessarily straightforward for targets that are not well studied. The target prediction databases can list thousands of targets for a particular miRNA, and they may not use standard nomenclature for the gene names of the targets, therefore it is necessary to search the target list with all known synonyms of the gene. In these cases it is useful to look up the gene or protein in NCBI or UniProt to find all possible synonyms – this can be very time-consuming and in some cases it is not possible to find a prediction, even though the authors have observed an effect on the levels of the mRNA in response to the miRNA. In these situations, it is recommended to annotate the miRNA with “negative regulation of gene expression”, with no indication of the target in the annotation extension, as described in Figure 7.

Finding an appropriate database identifier for a miRNA

Further information on the nomenclature of miRNAs can be found in miRBase (http://www.mirbase.org/help/nomenclature.shtml). The information available in a paper regarding the identification of a specific miRNA can be quite varied. Ideally there would be a sequence given that can then be used to search directly in RNAcentral (http://rnacentral.org/sequence-search/) for the appropriate identifier. Note: currently RNAcentral only provides sequence-specific identifiers, therefore there may be multiple identifiers for a given miRNA. You may have to reverse the sequence as some authors report the miRNA sequence in the 3' to 5' direction. On some occasions, an author may describe the study of a particular miRNA, for example miR-133, but the miRNA databases describe only miR-133a and miR-133b. In these cases where there is no evidence that the authors are describing a specific miRNA sequence, e.g. by providing a sequence, then unfortunately these miRNAs cannot be annotated.

Additionally, when a paper refers to a miRNA by its number, e.g. miR-121, without any sequence information, it is assumed that they are referring to the predominant product (rather than the product from the opposite arm of the precursor, indicated as e.g. miR-121*). If there is insufficient evidence to determine which is the predominant product, e.g. this is not indicated in the paper or there is no sequence shown, then the miRNA cannot be annotated, as the sequences are different.

Summary

When annotating a miRNA we should minimally aim to capture the following:

1. The miRNA’s main role in gene silencing and its target(s):

• “gene silencing by miRNA”/”negative regulation of gene expression” and “mRNA 3'UTR binding” and “mRNA binding involved in post-transcriptional gene silencing by miRNA”, if appropriate, with its target mRNA indicated in the annotation extension

2. The effect of silencing the target mRNA:

• e.g. if a cytokine transcript was silenced: “negative regulation of cytokine production”

3. The relevance of that process it regulates:

• e.g. “negative regulation of cytokine production involved in inflammatory process”.

Contributions

The miRNA guidelines were initially discussed at the Texas GO Consortium meeting in March 2014 involving the following people; Ruth Lovering, Rama Balakrishnan, Tanya Berardini, Judith Blake, Karen Christie, Harold Drabkin, Pascale Gaudet, David Hill, Doug Howe, Rachael Huntley, Donghui Li, Dmitry Sitnikov, Jennifer Smith, Kimberly Van Auken and Valerie Wood.

This manual was created by Rachael Huntley with the assistance of Ruth Lovering, Leonore Reiser and Dmitry Sitnikov. Acknowledgement goes to Peter D’Eustachio, Marc Gillespie, Anastasia Kalea, Simon Kay, Lars Maegdefessel, Lisa Matthews, Bruce May, Manuel Mayr, Marija Orlic-Milacic, Kimberly Van Auken and Anna Zampetaki for additional comments and suggestions.

References

Altuvia, Landgraf, Lithwick, Elefant, Pfeffer, Aravin, Brownstein, Tuschl and Margalit. 2005. "Clustering and conservation patterns of human microRNAs." Nucleic Acids Research. 33(8): 2697-706. doi: 10.1093/nar/gki567

Axtell, Michael J, Jakub O Westholm, and Eric C Lai. 2011. “Vive La Différence: Biogenesis and Evolution of microRNAs in Plants and Animals.” Genome Biology 12 (4): 221. doi:10.1186/gb-2011-12-4-221.

Bassett et al., 2014. "Understanding functional miRNA-target interactions in vivo by site-specific genome engineering." Nat Commun. 5:4640. doi: 10.1038/ncomms5640.

Boettger, Thomas, Nadine Beetz, Sawa Kostin, Johanna Schneider, Marcus Krüger, Lutz Hein, and Thomas Braun. 2009. “Acquisition of the Contractile Phenotype by Murine Arterial Smooth Muscle Cells Depends on the Mir143/145 Gene Cluster.” The Journal of Clinical Investigation 119 (9). American Society for Clinical Investigation: 2634–47. doi:10.1172/JCI38864.

Brock, Trenkmann, Gay, Michel, Gay, Fischler, Ulrich, Speich, and Huber. 2009. "Interleukin-6 modulates the expression of the bone morphogenic protein receptor type II through a novel STAT3-microRNA cluster 17/92 pathway." Circulation Research 104(10): 1184-91. doi: 10.1161/CIRCRESAHA.109.197491.

Castaldi, Alessandra, Tania Zaglia, Vittoria Di Mauro, Pierluigi Carullo, Giacomo Viggiani, Giulia Borile, Barbara Di Stefano, et al. 2014. “MicroRNA-133 Modulates the β1-Adrenergic Receptor Transduction Cascade.” Circulation Research 115: 273–83. doi:10.1161/CIRCRESAHA.115.303252.

Chen, Jian-Fu, Elizabeth M Mandel, J Michael Thomson, Qiulian Wu, Thomas E Callis, Scott M Hammond, Frank L Conlon, and Da-Zhi Wang. 2006. “The Role of microRNA-1 and microRNA-133 in Skeletal Muscle Proliferation and Differentiation.” Nature Genetics 38 (2): 228–33. doi:10.1038/ng1725.

Chi SW, Zang JB, Mele A, Darnell RB. 2009. "Argonaute HITS-CLIP decodes microRNA-mRNA interaction maps." Nature. 460(7254):479-86. doi: 10.1038/nature08170.

Chou et al. 2015. "miRTarBase 2016: updates to the experimentally validated miRNA-target interactions database". Nucleic Acids Research 44(D1):D239-D247. doi: 10.1093/nar/gkv1258

Cipolla, Gabriel a. 2014. “A Non-Canonical Landscape of the microRNA System” 5 (September): 1–6. doi:10.3389/fgene.2014.00337.

Cullen, BR. 2004. "Transcription and processing of human microRNA precursors." Molecular Cell. 16(6): 861-5. DOI: http://dx.doi.org/10.1016/j.molcel.2004.12.002

Dang, Truong-Minh, Wing-Cheong Wong, Siew-Min Ong, Peng Li, Josephine Lum, Jinmiao Chen, Michael Poidinger, Francesca Zolezzi, and Siew-Cheng Wong. 2015. “MicroRNA Expression Profiling of Human Blood Monocyte Subsets Highlights Functional Differences.” Immunology, February. doi:10.1111/imm.12456.

Ebert, Margaret S, Joel R Neilson, and Phillip A Sharp. 2007. “MicroRNA Sponges: Competitive Inhibitors of Small RNAs in Mammalian Cells.” Nature Methods 4 (9): 721–26. doi:10.1038/nmeth1079.

Fang Z, Rajewsky N. "The impact of miRNA target sites in coding sequences and in 3'UTRs." 2011. PLoS One. 2011 Mar 22;6(3):e18067. doi: 10.1371/journal.pone.0018067.

German MA, Pillay M, Jeong DH, Hetawal A, Luo S, Janardhanan P, Kannan V, Rymarquis LA, Nobuta K, German R, De Paoli E, Lu C, Schroth G, Meyers BC, Green PJ. 2008. "Global identification of microRNA-target RNA pairs by parallel analysis of RNA ends." Nat Biotechnol. 26(8):941-6. doi: 10.1038/nbt1417.

Guo, Huili, Nicholas T Ingolia, Jonathan S Weissman, and David P Bartel. 2010. “Mammalian microRNAs Predominantly Act to Decrease Target mRNA Levels.” Nature 466 (7308): 835–40. doi:10.1038/nature09267.

Grosswendt S, Filipchyk A, Manzano M, Klironomos F, Schilling M, Herzog M, Gottwein E, Rajewsky N. 2014. "Unambiguous identification of miRNA:target site interactions by different types of ligation reactions." Mol Cell. 54(6):1042-54. doi: 10.1016/j.molcel.2014.03.049.

Hafner M, Landthaler M, Burger L, Khorshid M, Hausser J, Berninger P, Rothballer A, Ascano M Jr, Jungkamp AC, Munschauer M, Ulrich A, Wardle GS, Dewell S, Zavolan M, Tuschl T. 2010. "Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP." Cell. 141(1):129-41. doi: 10.1016/j.cell.2010.03.009.

Hausser, Jean, Afzal Pasha Syed, Biter Bilen, and Mihaela Zavolan. 2013. “Analysis of CDS-Located miRNA Target Sites Suggests That They Can Effectively Inhibit Translation.” Genome Research 23 (4): 604–15. doi:10.1101/gr.139758.112.

Helwak and Tollervey. 2014. "Mapping the miRNA interactome by crosslinking ligation and sequencing of hybrids (CLASH)" Nature protocols. 9(3):711-728. doi:10.1038/nprot.2014.043.

Huntley RP et al. 2016. "Guidelines for the functional annotation of microRNAs using the Gene Ontology." RNA 22: 667-676. doi: 10.1261/rna.055301.115.

Kuhn et al. 2008. "Experimental Validation of miRNA Targets". Methods. 44(1): 47–54.

Landgraf P1, Rusu M, Sheridan R, Sewer A, Iovino N, Aravin A, Pfeffer S, Rice A, Kamphorst AO, et al. 2007. "A mammalian microRNA expression atlas based on small RNA library sequencing." Cell. 129(7):1401-14. doi:10.1016/j.cell.2007.04.040.

Lauressergues, Dominique, Jean-Malo Couzigou, Hélène San Clemente, Yves Martinez, Christophe Dunand, Guillaume Bécard, and Jean-Philippe Combier. 2015. “Primary Transcripts of microRNAs Encode Regulatory Peptides.” Nature 520 (7545): 90–93. doi:10.1038/nature14346.

Nicolas, Francisco E. 2011. “Experimental Validation of microRNA Targets Using a Luciferase Reporter System.” Methods in Molecular Biology (Clifton, N.J.) 732 (January): 139–52. doi:10.1007/978-1-61779-083-6_11.

Orom UA, Lund AH. 2007. "Isolation of microRNA targets using biotinylated synthetic microRNAs." Methods. 43(2):162-5. doi:10.1016/j.ymeth.2007.04.007

Rogers, Kestrel, and Xuemei Chen. 2013. “Biogenesis, Turnover, and Mode of Action of Plant microRNAs.” The Plant Cell 25 (7): 2383–99. doi:10.1105/tpc.113.113159.

Sayed, Danish, Shweta Rane, Jacqueline Lypowy, Minzhen He, Ieng-Yi Chen, Himanshu Vashistha, Lin Yan, Ashwani Malhotra, Dorothy Vatner, and Maha Abdellatif. 2008. “MicroRNA-21 Targets Sprouty2 and Promotes Cellular Outgrowths.” Molecular Biology of the Cell 19 (8): 3272–82. doi:10.1091/mbc.E08-02-0159.

Selbach M, Schwanhäusser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N. 2008. "Widespread changes in protein synthesis induced by microRNAs." Nature. 455(7209):58-63. doi: 10.1038/nature07228.

Steele, R., J. L. Mott, and R. B. Ray. 2010. “MBP-1 Upregulates miR-29b, Which Represses Mcl-1, Collagens, and Matrix Metalloproteinase-2 in Prostate Cancer Cells.” Genes & Cancer 1 (4): 381–87. doi:10.1177/1947601910371978.

Stroynowska-Czerwinska, Anna, Agnieszka Fiszer, and Wlodzimierz J. Krzyzosiak. 2014. “The Panorama of miRNA-Mediated Mechanisms in Mammalian Cells.” Cellular and Molecular Life Sciences 71: 2253–70. doi:10.1007/s00018-013-1551-6.

Sun, Gao, Zhou, Wang, Zhang, Deng and Li. 2013. "Comparative genomic analysis reveals evolutionary characteristics and patterns of microRNA clusters in vertebrates." Gene. 512(2): 383-91. doi: 10.1016/j.gene.2012.09.102.

Thomson, Bracken and Goodall. 2011. "Experimental strategies for microRNA target identification" Nucleic Acids Research 39(16): 6845-6853. doi: 10.1093/nar/gkr330.

Van Rooij, Eva, William S Marshall, and Eric N Olson. 2008. “Toward microRNA-Based Therapeutics for Heart Disease: The Sense in Antisense.” Circulation Research 103 (9): 919–28. doi:10.1161/CIRCRESAHA.108.183426.

Vanhecke, Dominique, and Michal Janitz. 2004. “High-Throughput Gene Silencing Using Cell Arrays.” Oncogene 23 (51): 8353–58. doi:10.1038/sj.onc.1208027.

Vo NK, Dalton RP, Liu N, Olson EN, Goodman RH."Affinity purification of microRNA-133a with the cardiac transcription factor, Hand2." Proc Natl Acad Sci U S A. 107(45):19231-6. doi: 10.1073/pnas.1013162107.

Wakiyama, Motoaki, Koji Takimoto, Osamu Ohara, and Shigeyuki Yokoyama. 2007. “Let-7 microRNA-Mediated mRNA Deadenylation and Translational Repression in a Mammalian Cell-Free System.” Genes & Development 21 (15): 1857–62. doi:10.1101/gad.1566707.

Wang, Jing-Jing, and Hui-Shan Guo. 2015. “Cleavage of INDOLE-3-ACETIC ACID INDUCIBLE28 mRNA by MicroRNA847 Upregulates Auxin Signaling to Modulate Cell Proliferation and Lateral Organ Growth in Arabidopsis.” The Plant Cell 27 (3): 574–90. doi:10.1105/tpc.15.00101.

Wang, Shusheng, Arin B Aurora, Brett A Johnson, Xiaoxia Qi, John McAnally, Joseph A Hill, James A Richardson, Rhonda Bassel-Duby, and Eric N Olson. 2008. “The Endothelial-Specific microRNA miR-126 Governs Vascular Integrity and Angiogenesis.” Developmental Cell 15 (2): 261–71. doi:10.1016/j.devcel.2008.07.002.

Wilczynska, a, and M Bushell. 2014. “The Complexity of miRNA-Mediated Repression.” Cell Death and Differentiation 22 (1). Nature Publishing Group: 22–33. doi:10.1038/cdd.2014.112.

Winter, Julia, Stephanie Jung, Sarina Keller, Richard I Gregory, and Sven Diederichs. 2009. “Many Roads to Maturity: microRNA Biogenesis Pathways and Their Regulation.” Nature Cell Biology 11 (3): 228–34. doi:10.1038/ncb0309-228.

Woods, Thomson and Hammond. 2007. "Direct regulation of an oncogenic micro-RNA cluster by E2F transcription factors." J. Biol. Chem. 282(4): 2130-4. doi:10.1074/jbc.C600252200.

Xie, Meng, Shuxin Zhang, and Bin Yu. 2015. “microRNA Biogenesis, Degradation and Activity in Plants.” Cellular and Molecular Life Sciences : CMLS 72 (1): 87–99. doi:10.1007/s00018-014-1728-7.

Young, Jennifer A, Ka Ka Ting, Jia Li, Thorleif Moller, Louise Dunn, Ying Lu, Joshua Moses, et al. 2013. “Regulation of Vascular Leak and Recovery from Ischemic Injury by General and VE-Cadherin-Restricted miRNA Antagonists of miR-27.” Blood 122 (16). American Society of Hematology: 2911–19. doi:10.1182/blood-2012-12-473017.

Yu, Wang, Yang, Wang, Ma, Du and Zhang. 2006. "Human microRNA clusters: genomic organization and expression profile in leukemia cell lines." Biochem Biophys Res Commun. 349(1): 59-68. doi:10.1016/j.bbrc.2006.07.207

Zampetaki, a., R. Attia, U. Mayr, R. S. M. Gomes, a. Phinikaridou, X. Yin, S. R. Langley, et al. 2014. “Role of miR-195 in Aortic Aneurysmal Disease.” Circulation Research 115: 857–66. doi:10.1161/CIRCRESAHA.115.304361.


Zeng Y, Pan Y, Liu H, Kang K, Wu Y, Hui G, Peng W, Ramchandran R, Raj JU, Gou D. 2014. "MiR-20a regulates the PRKG1 gene by targeting its coding region in pulmonary arterial smooth muscle cells." FEBS Lett. 2014 Dec 20;588(24):4677-85. doi: 10.1016/j.febslet.2014.10.040.


Zhang, Zhang and Su. 2009. "Diversity and evolution of MicroRNA gene clusters." Sci China C Life Sci. 52(3): 261-6. doi: 10.1007/s11427-009-0032-5.

Zhang, Yijun, Miaomiao Fan, Xue Zhang, Feng Huang, Kang Wu, Junsong Zhang, Jun Liu, et al. 2014. “Cellular microRNAs up-Regulate Transcription via Interaction with Promoter TATA-Box Motifs.” RNA (New York, N.Y.) 20 (12): 1878–89. doi:10.1261/rna.045633.114.

Zhao, Yong, Joshua F Ransom, Ankang Li, Vasanth Vedantham, Morgan von Drehle, Alecia N Muth, Takatoshi Tsuchihashi, Michael T McManus, Robert J Schwartz, and Deepak Srivastava. 2007. “Dysregulation of Cardiogenesis, Cardiac Conduction, and Cell Cycle in Mice Lacking miRNA-1-2.” Cell 129 (2): 303–17. doi:10.1016/j.cell.2007.03.030.

Zhu W, Yang L, Du Z. 2011. "MicroRNA Regulation and Tissue-Specific Protein Interaction Network". PLoS ONE, 6(9): e25394. doi:10.1371/annotation/60addd69-bd2d-4ecf-8369-6a5a1bf2cd04.