Elucidating miRNA expression in the inner ear
The initial discovery of miRNAs in the inner ear was based on their expression (Table 1). Since then, a number of methods have been used to explore the temporal and spatial expression of miRNAs. Microarrays have generally been used for the study of gene expression profiles. They have been especially valuable for expression profiles of small RNAs, including miRNAs (Barad et al, 2004; Liu et al, 2004). This technique compares between the expression profiles of two samples, using hybridization of each sample to oligonucleotide-synthesized labelled probes. The first mammalian inner ear miRNAs were described following hybridization of whole cochlea at different time points to commercial miRNA microarrays (Weston et al, 2006). Affymetrix microarrays have been used for gene expression analysis between proliferating or non-proliferating chick auditory epithelia (Frucht et al, 2010). Thereafter, locked nucleic acid (LNA)-enhanced arrays were used, based on the nucleic acid analogue LNA (Petersen & Wengel, 2003) with high affinity to RNA (Elkan-Miller et al, 2011). This microarray platform, manufactured by Exiqon A/S (Denmark), is highly sensitive and specific, with both known and proprietary miRNAs.
Quantitative real-time polymerase chain reaction (qRT-PCR) is a rapid and sensitive method to quantify mature or precursor miRNAs. They are often used following identification of miRNA expression with microarrays, to both validate the finding and define the differential levels of expression accurately. This type of validation of miRNAs identified in the first mammalian microarray screen utilized SYBR Green fluorescence (Weston et al, 2006). This method was used further to examine changes in miRNA expression in cochlear progenitor cells (Hei et al, 2011). Other validations utilized the TaqMan system, which is based on stem and loop primers to convert RNA to cDNA, increasing the specificity and solving the problem of similarity of miRNA family members differing by only one nucleotide (Chen et al, 2005; Raymond et al, 2005). In a recent report comparing several expression profile methods for detection of miRNAs in a cancer cell line, there appeared to be a high correlation between the different technical methods (Sokilde et al, 2011).
In situ hybridization is widely used to characterize the spatial expression profile of miRNAs. The tissue, either in whole-mount form or in frozen or paraffin sections, is hybridized to an LNA™ modified detection probe, labelled with a digoxygenin (DIG) molecule, and followed by detection with an anti-alkaline phosphatase-conjugated (AP) antibody (Weston et al, 2006). The result is a purple staining where the miRNA is detected, after reacting with the alkaline phosphatase substrate. The most striking results using this technique have demonstrated that miRNAs have an extremely variable temporal and spatial expression pattern (Elkan-Miller et al, 2011; Friedman et al, 2009; Sacheli et al, 2009).
Due to the inaccessibility of human inner ear tissue and lack of relevant human inner ear cell lines, expression studies on miRNAs are limited to animals, with extrapolations to human function. For these reasons both the zebrafish and mouse are organisms of choice for evaluating ear miRNA function. Zebrafish are commonly used as a powerful tool to study mammalian systems and are a relevant model to study human disease (Driever & Fishman, 1996). These fish are small, easy to grow, translucent and are relatively easy to manipulate by gene silencing at the embryonic stage. The perception of hearing and balance in the zebrafish is performed by the lateral line and the mechanosensory organs (Nicolson, 2005). miRNA expression in the zebrafish inner ear was first demonstrated by mir-200a and mir-183 expression in the sensory epithelia (Wienholds et al, 2005). A conserved miRNA cluster, which includes miR-183, miR-182 and miR-96, was shown to be expressed in the zebrafish in the hair cells, otic neurons and other primary sensory cells. These findings were followed by a report that miR-15a-1 is found in neuromasts and throughout the inner ear and miR-18a is mainly in the utricular macula and nearby cells at 48 h post-fertilization (hpf) (Friedman et al, 2009). As some of these miRNAs were found to be restricted to the sensory organs, a role for miRNA regulation was implicated in this complex system.
The mouse is considered to be the most relevant mammalian model to study mechanisms of deafness (Leibovici et al, 2008). Inner ear development in the mouse is an extensively researched field, and numerous deaf mouse mutants serving as models for human deafness have been studied (Frolenkov et al, 2004). The emerging interest with miRNA expression in the mouse inner ear and their connection to development and maturation of the inner ear began with a microarray expression analyses study throughout several developmental stages, from postnatal day zero (P0) to 5 weeks (Weston et al, 2006). A subset of miRNAs were then evaluated by Northern blot analysis using samples from the inner ear and other mouse organs, to assess whether there are tissue-specific miRNAs, with a special interest in the inner ear. These analyses revealed that the conserved cluster of mir-183, mir-182 and mir-96 have a restricted expression to the mouse inner ear, as compared to brain, heart and whole embryo expression. In situ hybridization revealed the unique expression pattern of mir-182, mir-183 and mir-96 in inner and outer hair cells of the cochlea, hair cells of the vestibular organs and spiral and vestibular ganglia. These three miRNAs were shown to be expressed in the same pattern (Weston et al, 2006), are transcribed in the same orientation and are predicted to be co-expressed (Wienholds et al, 2005). Another distinctive expression pattern was exhibited by miR-124, an abundantly expressed miRNA in the nervous system (Lagos-Quintana et al, 2002), which was found to be expressed in spiral and vestibular ganglia (Weston et al, 2006). This was the first report describing mammalian inner ear specific miRNAs, and their ubiquitous or restricted expression patterns.
In order to address the essential question of the effect of miRNAs throughout development, a study was conducted examining the expression pattern of the mir-183, mir-182 and mir-96 cluster (Sacheli et al, 2009). In this study, in situ hybridization analysis was performed at early time points, with miRNAs displaying distinctive developmental expression. The earliest observed expression was at embryonic day 9.5 (E9.5). miR-183 and miR-182, but not miR-96, were detected in the otic vesicle in the embryonic early inner ear. Later, at E11.5, all three miRNAs were detected in the otic vesicle, in the cochlea-vestibular ganglion, and the neural tube. The dramatic change in expression begins around E14.5, as at this time point this cluster's expression is limited to the vestibular hair cells. From E17.5 expression was detected in the hair cells of what will become the cochlea and the vestibular system, and were not detected in the supporting cells. By P0, miR-183, miR-182 and miR-96 were strongly expressed in hair cells of the cochlea and the vestibular system, and in the spiral ganglia. At P4 and on, there appeared to be another big shift in expression, as miR-96 was no longer observed in the cochlea or vestibule, but was detected in the spiral limbus and the inner sulcus. However, the expression of miR-183 and miR-182 continued in the hair cells, but ceased to be present in hair cells from P11-15, and was only found in the spiral limbus and the inner sulcus. Overall, these findings correlate with the maturation course of inner ear development and differentiation (Fig 1). The differentiation of the hair cells begins around E14.5, until around E17-E18, when the sensory epithelium exhibit hair cells throughout its length (Chen et al, 2002). The fact that expression of these miRNAs peak at the point of differentiation, and change from a general expression pattern in the sensory epithelium to restricted expression in the hair cells implicates these miRNAs in the development and maturation of the neonatal mouse inner ear.
Figure 1. Timeline for miRNA expression during development and early postnatal stages of the inner ear for a subset of miRNAs. The most well-studied miRNAs, mir-96, -182- and -183, were first detected in the otic vesicle at E9.5, progressed to the cochlea-vestibule ganglion and the neural tube at E11.5, and by E17.5 had strong expression in the cochlear hair cells. The expression continued to at least p30, by some reports. The expression of other miRNAs detected by in situ hybridization are shown as well. Those marked with # were only examined at the stage indicated.
Download figure to PowerPoint
Another study with an interest in the regulatory differences between the cochlea and the vestibular systems carried out a microarray scan of small RNAs from mouse P0 cochlea and vestibule (Friedman et al, 2009). Twenty-four miRNAs were found to be differentially expressed between the two tissues. An expression profile of a set of five miRNAs was evaluated using in situ hybridization. mir-15a, mir-18a, mir-30b, mir-99a and mir-199a were found in different and distinct regions of the mouse P0 cochlea and vestibule, including hair and supporting cells, the spiral ganglia and other cell types. The different expression patterns demonstrated that miRNA expression in the inner ear is not restricted to one cell type or developmental stage, and their expression is spread and is associated with the developmental changes that occur during the course of time (Fig 1).
miRNAs are involved in formation of the inner ear
Once it was established that miRNAs are expressed extensively throughout inner ear and during development, the involvement of Dicer1 in the formation of the inner ear was addressed as well. Type III ribonuclease Dicer1 is essential to the processing of mature miRNAs from their pre-miRNA form (Bernstein et al, 2001). Mice lacking the Dicer1 enzyme, created by a knock-out (KO) gene deletion technique, die during embryogenesis, as observed as early as E7.5, probably due to the inability of the embryo to produce functional mature miRNAs that may regulate developmental processes (Bernstein et al, 2003).
To overcome this early death, conditional knock-outs were created using the Cre transgene (Harfe et al, 2005). The first study investigating the role of Dicer1 in the development of the inner ear was conducted using Pax2-Cre transgenic mice (Soukup et al, 2009). The conditional Dicer1 deletion in the Pax2-Cre mice allows for Dicer1 deletion in the inner ear, kidney and midbrain upon crossing to the Dicer1-loxP mice. The dramatic influence of the loss of Dicer1 was demonstrated by the mice dying by E18.5. The inner ear malformations were extensive. By E17.5 there was a major truncation of the inner ear structure, and the cochlea was missing the coiled shape. Scanning electron microscopy (SEM) revealed that cochlear hair cells exhibited abnormal stereocilia organization. A different conditional Dicer1 KO model used Cre expression downstream to the Pou4f3 promoter in an attempt to create viable mice (Dicer-PCKO) (Friedman et al, 2009). The Pou4f3 promoter was chosen for the Dicer1 conditional KO mouse model as it is expressed in the hair cells. Restricting the expression of Cre recombinase in these cells resulted in the removal of the Dicer1-LoxP sites from the hair cells, leading to depletion of Dicer1 expression in hair cells. The majority of defects in the auditory system were observed in the adult mice. Dicer-PCKO p38 mice were deaf, as was demonstrated by auditory brainstem response (ABR), and also showed a mild circling vestibular phenotype. Most of the cochlear hair cells were misshaped and did not express myosin VI, a hair cell marker and SEM revealed that many of the hair cells either lost their stereocilia or had abnormal shapes. These results demonstrate a more profound inner ear malformation achieved by interrupting the miRNA maturation process. The next model used a Foxg1-Cre mediated knock out of Dicer1 (Kersigo et al, 2011). The effect on the inner ear was significant, with reduction in the inner ear size observed at E14.5, followed by a loss of canal formation and a loss of the coiled cochlea structure by E18.5. These models indicate that normal development of the functioning inner ear is strongly dependent on normal miRNA maturation. Disruption of the mature miRNA production caused not only structural defects, but a loss of auditory and vestibular function.
The last model used so far to investigate the influence of Dicer1 on inner ear development is a Atoh1-Cre conditional Dicer1 knockout. Atoh1 is expressed in all hair cells during the embryonic stage, but is not restricted to the inner ear. Mice died around the age of 4 weeks, after exhibiting ataxia and seizures. miR-183 was completely depleted in the hair cells of the mutant mice, but not in spiral ganglia. In addition, expression of miRNAs was observed in other inner ear components, but not in hair cells, indicating that this model specifically suppresses the expression of miRNA in hair cells (Weston et al, 2011). These findings suggest that miRNAs are crucial for inner ear development, and are involved in the formation, morphogenesis and neurosensory processes that create the functional auditory organ.
Approaches to miRNA target identification
In order to understand how miRNAs function in cells, their targets must be identified. miRNAs destabilize target mRNAs or inhibit protein translation, leading to decreased protein levels (Guo et al, 2010). These targets may be important cell regulators and their decrease, even by a small amount, can change the course of the cell. Target prediction has been a challenge since the discovery of miRNAs; however, algorithms to find real functional targets are improving. There are a number of publicly available target prediction programs, including but not limited to TargetScan and miRanda, most of which work by slightly different algorithms, although they all rely on seed pairing of miRNA-target recognition. They search for the complementation of the miRNA 5′ seed region nucleotide (nt) 2–7 to the 3′UTR of the target, and the evolutionary conservation of the 7 nt in 3′UTR of targets, matching the miRNA seed region (Bartel, 2009). The significant shortcoming of these computational prediction programs is the large amount of false positive noise they produce. For example, although a miRNA-target pair could be found using such a method, both the miRNA and target are not necessarily expressed in the studied tissue and thus do not interact in vivo. In such a case, the prediction is not relevant. This conclusion emphasizes the importance of biological validation of each computationally predicted miRNA-target pair. Identification of new targets for miRNAs expressed in the inner ear has been demonstrated in a few studies (Elkan-Miller et al, 2011; Hertzano et al, 2011).
A computational analysis using the miRanda target prediction tool was conducted to produce a list of more than 100 putative targets for miR-96 (Lewis et al, 2009). The list was annotated and 13 potential targets were then examined biologically using a luciferase assay (Lewis et al, 2009). Five of the tested targets were validated biologically as miR-96 targets: Aqp5, Celsr2, Myrip, Odf2 and Ryk. Using the miR-96 mouse mutant diminuendo, a different target identification approach was used. Microarray gene expression analysis was done to compare the differentially expressed set of mRNAs from the mutant diminuendo mouse and a wild type (WT) control. After qRT-PCR confirmation, five genes were found to be down-regulated in the mutant: Slc26a5 (prestin), Ocm (oncomodulin), Pitpnm1, Gfi1 and Ptprq. These genes do not appear to be direct targets of miR-96, suggesting that their down-regulation is a downstream effect.
Another study used a combination of experimental data and computational analysis to predict functional targets for miRNAs in the inner ear (Elkan-Miller et al, 2011). miRNAs were identified after applying a transcriptomic and proteomic expression profiling to search for differentially expressed miRNAs between the cochlea and vestibular P2 mouse sensory epithelia. Functional targets were then searched for by using the FAME (Functional Assignment of miRNA via Enrichment) algorithm (Ulitsky et al, 2010). FAME evaluates enrichment and depletion of potential targets, found by TargetScan, in a data set of up or down-regulated mRNAs and proteins. In this study, miR-135b was found in LNA arrays to be up-regulated in the vestibule, while its targets were enriched in a protein dataset and down-regulated in the vestibule. In situ hybridization confirmed the differential expression of miR-135b in vestibular hair cells as compared to cochlear hair cells at P0. The analysis predicted PSIP1-P75 as one of potential targets of miR-135b. Psip, PC4- and SF-2 interacting protein/Ledgf is a known transcriptional activator implicated to regulate stress-related genes, has an anti-apoptotic effect, is involved in mRNA splicing, cell survival and is part of a fusion gene in leukaemia (Sutherland et al, 2006). Both qRT-PCR and protein expression was consistent with the FAME predictions, as PSIP1-P75 was down-regulated in the vestibule, as compared to the cochlea. Finally, the miRNA-target interaction was verified by both an RNA interference (RNAi) silencing technique and a luciferase reporter assay. This example illustrates a target identification approach, combining both in silico analysis and experimental data to be able to predict functionally relevant targets. Furthermore, the pathways involving PSIP1-P75 are known in other tissues and can provide guidelines for determining the function of this miRNA-target pair in the inner ear (Fig 2).
Figure 2. miRNA target identification. Computational and experimental approaches have been taken to identify target of miRNA in the inner ear. An example of such a miR-target pair is miR-135 and Psip, PC4- and SF-2 interacting protein/Ledgf (Elkan-Miller et al, 2011), which has been implicated in transcriptional regulation of stress-related genes, having an anti-apoptotic effect, involved in mRNA splicing, cell survival and is part of a fusion gene in leukaemia.
miR-135 is reduced in the cochlear hair cells, while its expression is high in vestibular hair cells.
Psip1, one of its targets, is expressed in the nucleus of the hair cells.
The pathways shown demonstrate potential inner ear functional pathways implicated in the miR135b-Psip regulatory network. Psip1 is a transcriptional regulator that plays a role in retinoic acid production (Fatma et al, 2004
). Retinoic acid is crucial for hair cell development (Raz & Kelley, 1999
). This model suggests that miR-135b regulation of Psip1 plays a role in hair cell development and survival.
Download figure to PowerPoint
The final example of a complex approach to search for functional targets was demonstrated by a targeted search of inner ear regulators of cell fate, including miRNA targets and transcription factors, using a transcriptome cell type-specific study (Hertzano et al, 2011). Flow cytometry of vestibular and cochlear tissue using CD326, CD34 and CD49f markers allowed separation of epithelial (sensory and nonsensory), neuronal and vascular cell populations to identify differentially expressed transcripts between these eight distinct cell populations. Using this data, miRNA families whose potential targets were differentially expressed in a specific cell type, in the opposite direction of the miRNA expression, were coupled. The prediction identified the miR-200b family, with potential targets down-regulated in sensory epithelial cells at P0. miR-200b expression was detected by in situ hybridization in all sensory epithelial cells of the inner ear. In order to identify other regulatory factors in the inner ear, a search was made for cis-regulatory elements over expressed in promoters of sensory epithelial cells in the inner ear. This bioinformatics technique revealed the transcription factors ZEB1 and ZEB2 as regulators of genes expressed in the inner ear sensory cells. It has been previously shown that miR-200b directly targets ZEB1 and ZEB2 (Burk et al, 2008), linking the regulatory factors to one another. Finally, using a cell type-specific approach, they demonstrated marked mis-expression of genes suggested to be regulated by Zeb1 in a mouse mutant with a mutation in the Zeb1 transcription factor. The results of this study uncovered a complex regulatory network of miRNAs, targets and transcriptional repressors.