A mouse model of miR-96, miR-182 and miR-183 misexpression implicates miRNAs in cochlear cell fate and homeostasis

Germline mutations in Mir96, one of three co-expressed polycistronic miRNA genes (Mir96, Mir182, Mir183), cause hereditary hearing loss in humans and mice. Transgenic FVB/NCrl- Tg(GFAP-Mir183,Mir96,Mir182)MDW1 mice (Tg1MDW), which overexpress this neurosensory-specific miRNA cluster in the inner ear, were developed as a model system to identify, in the aggregate, target genes and biologic processes regulated by the miR-183 cluster. Histological assessments demonstrate Tg1MDW/1MDW homozygotes have a modest increase in cochlear inner hair cells (IHCs). Affymetrix mRNA microarray data analysis revealed that downregulated genes in P5 Tg1MDW/1MDW cochlea are statistically enriched for evolutionarily conserved predicted miR-96, miR-182 or miR-183 target sites. ABR and DPOAE tests from 18 days to 3 months of age revealed that Tg1MDW/1MDW homozygotes develop progressive neurosensory hearing loss that correlates with histologic assessments showing massive losses of both IHCs and outer hair cells (OHCs). This mammalian miRNA misexpression model demonstrates a potency and specificity of cochlear homeostasis for one of the dozens of endogenously co-expressed, evolutionally conserved, small non-protein coding miRNA families. It should be a valuable tool to predict and elucidate miRNA-regulated genes and integrated functional gene expression networks that significantly influence neurosensory cell differentiation, maturation and homeostasis.

in the case of HC cell fate assignment, then in a fine-tuning manner in the case of HC morpho-functional maturation and homeostasis.
During development of the cochlear neurosensory epithelium (i.e. organ of Corti or OC), the prosensory cell lineage differentiates as either HC or supporting cells (SCs), a process mediated in part through Delta-Notch signaling 20 . By E15.5, miR-183 cluster expression distinguishes differentiated cochlear HC, which suggests a role for these miRNAs in this cell type transition 9 . Based on these data, we hypothesized that the miR-183 cluster, if misexpressed in SCs, would, in the context of mutual exclusion, perturb SC gene expression. The observation  of aggregate effects would help establish evolutionarily conserved miRNA function(s) with potential relevance with regard to the treatment of hearing loss through HC regeneration strategies 6 . To test this hypothesis, we engineered Tg(GFAP-Mir183,Mir96,Mir182) (Tg 1MDW ) mice to drive ectopic miR-183 cluster expression using the core human promoter of the glial fibrillary acidic protein (GFAP). Overall, Tg 1MDW cochlear HC and SC differentiate and histologically mature in a similar pattern to that of wild-type littermates. Nonetheless, HC lifespan is significantly shortened and hearing function is rapidly lost: most cochlear HCs are lost by the time the mice are 6 months old, resulting in severe deafness. These data are consistent with the view that HC microRNA misexpression in lineage-related SCs has effects on their capacity to maintain HC homeostasis. Furthermore, the gene expression effects, in aggregate and in identified gene ontology pathways, provide basic information necessary to gain a more complete understanding not only of the role these microRNAs serve during HC maturation, but also those genes and processes that are particularly sensitive to SC miRNA misexpression.  24 . Both homozygous and hemizygous mice from Lines Tg 1MDW and Tg 2MDW develop lens cataracts with average onset ages, in days, of 95 ± 11 n = 127, 121 ± 23 n = 19, 132 ± 17 n = 12, 157 ± 5 n = 14, respectively (Kruskal-Wallis P < 0.0001). The Tg 3MDW homozygous and hemizygous mice exhibit a progressive loss of pelage clearly evident by P50 that progresses steadily until these mice become hairless. The phenotypes are likely due to effects on expression patterns and levels due to transgene copy number and integration site differences (Table 1). Indeed, initial characterization of the three lines in brain tissue showed substantial differences in miR-183 cluster miRNA abundance by RT-PCR (Tg 1MDW >>> Tg 2MDW > Tg 3MDW , Supplementary Figure 1A). Also, in-situ hybridization (ISH) with an LNA-DIG labeled probe against miR-182 was consistent with RT-PCR, showing Bergmann glial cell localization in Tg 1MDW cerebellum only, a common cell type expressing GFAP-core promoter transgenes 23 (Supplementary Figure 1B- 11,12 . Interestingly, this extra IHC patterning in Tg 1MDW/1MDW phenocopies both hypomorphic and null mutants of genes (i.e. Jag1, Sox2, Hes1) that are: 1) markers of differentiated SCs and; 2) predicted targets of miR-183 cluster miRNAs (Fig. 2D) [25][26][27][28][29] . For Jag1 +/− , the extra IHCs estimate was extrapolated from counts of adult basal cochleae of headturner (Jag1 +/Htu ) and slalom (Jag1 +/Slm ) heterozygous mice 26,27 . For Sox2, extra IHCs were similarly extrapolated from counts taken from the entire length of late embryonic (E18) Sox2 +/EGFP reporter knockout (KO) and Sox EGFP/LP hypomorphic cochleae 28 . For Hes1, counts were from 1 mm lengths starting from the basal end of E18 cochleae of KO Hes1 +/− and Hes1 −/− mice 25 . While the mechanism for the genesis of these extra IHCs is unknown, one possibility is that the transgenic miR-183 cluster expression promotes SC-to-HC transdifferentiation. In support of this idea, we detected a discrete number of apparent Deiters' cells positive for both MYO6 and miR-182 at P18 (Fig. 3A,B, arrows). The cells appear to be histomorphologically hybrid in nature by: 1) showing decreased MYO6 staining compared to HCs; 2) having endfeet in contact with the basilar membrane, a characteristic of all OC SCs and; 3) displaying nuclei located at the same z-plane as OHC nuclei.  Luciferase activity for each experimental replicate was normalized to that of the scrambled control siRNA transfected replicate (CTRL) (Fig. 4). As expected, relative luciferase activity in cells co-transfected with miR-182 and reporter vector containing the Sox2 3′ UTR is reduced by nearly 20%, confirming previous findings that Sox2 is a miR-183 cluster target 9 . Luciferase activity is significantly repressed in cells co-transfected with synthetic miR-96 or miR-183 and pmirGLO-Notch1 3′ UTR, suggesting that these miR-183 cluster members directly interact with and silence Notch1. Similarly, Hes1 validates as a target of miR-96 and miR-182, with miR-96 downregulation of luciferase activity exceeding 20%. Jag1, however, failed to validate as a target of any miR-183 cluster member with statistical significance. This is likely due to G:U wobble base pairs at nucleotides 4 and 5 of miR-183 which, while thermodynamically favorable, can drastically reduce the efficacy of miRNA-mediated translational repression 30 . Taken together, these results suggest that Sox2, Notch1, and Hes1 are genuine miR-183 cluster targets. Sox2 and Atoh1 exhibit a mutually antagonistic relationship in SC/HC differentiation 28 , while Notch1-regulated Hes1 gene expression also antagonizes the ability of Atoh1 to promote hair cell differentiation 31 . Given that miR-183 cluster expression requires Atoh1-mediated HC specification 9 , the miR-183 cluster likely serves a crucial function in HC differentiation by downregulating Sox2, Notch1, and Hes1, thus contributing to Atoh1 specific HC fate determination.
To provide a biologic context to these expression changes, we performed gene ontology (GO) analysis using the DAVID bioinformatics resources 32 . For this, we submitted gene lists that included changed genes down to a ± 1.5-fold change cutoff (FDR-adjusted P < 0.05). Pathway analysis revealed P5 downregulated genes were significantly enriched for genes involved in myelination and ion homeostasis; P18 downregulated genes were related to synapse structure; and P18 upregulated genes were related to muscle and the actin cytoskeleton (Table 4). To evaluate transcriptome effects directly attributable to the miR-183 cluster, we performed hypergeometric analyses using a non-redundant set of evolutionarily conserved miRNA/3′UTR seed sequences: 153 vertebrate miR-NAs and 62,793 3′UTR seed sites (8mer, 7mer-m8, 7mer-1A, 3comp, TargetScan 6.2 29 ). This seed site universe was used to test for enrichment/depletion within upregulated and downregulated gene sets. For each miR-183 cluster miRNA, we found enrichment (Odds Ratio 1.43-1.45, FDR-adjusted P < 0.05, Table 5) of evolutionarily conserved 3′UTR seed targets in P5 Tg 1MDW/1MDW downregulated genes (uncorrected P < 0.05, fold change <−1.05). Importantly, this enrichment is consistent with repression of miR-183 cluster target genes at evolutionarily conserved sites in Tg 1MDW/1MDW P5-OC. In P18-cochlea, seed sites of 23 and 6 vertebrate miRNAs were either enriched (Odd Ratio >1, FDR-adjusted P < 0.05) or depleted (Odds Ratio <1, FDR-adjusted P < 0.05) in P18-cochlea upregulated or downregulated gene sets, respectively (uncorrected P < 0.05, fold change <−1.05 or >+ 1.05, Table 5). Only miR-96 seed sites were found depleted in upregulated genes at P18, an effect opposite to that observed for this miRNA at P5-OC.
For validation of our methodology, we applied the same hypergeometric analysis of the TargetScan 6.2 seed site universe to published microarray data from P4-OC miR-96 Diminuendo mouse (miR96 ddl/ddl ) 2 and from 5wk-retina miR-183 cluster KO mouse (miR-183C GT/GT ) 33 . In the case of the miR96 ddl/ddl microarray, miR-96 seed sites were enriched (Odds Ratio 2.10, FDR-adjusted P < 0.05) in miR96 ddl/ddl upregulated genes (uncorrected P < 0.05, fold change >+ 1.05). This independently confirms specific de-repression of evolutionarily conserved miR-96 target genes due to a miR-96 single nucleotide mutation 2 . In miR-183C GT/GT 5wk-retina, our hypergeometric analysis demonstrates that combined loss of all miR-183 cluster miRNAs results not only in enrichment (Odds Ratio 1.65-1.70, FDR-adjusted P < 0.05, Table 5) of miR-183 cluster targets in miR-183C GT/GT upregulated  , and Hes1 as miR-183 cluster by dual luciferase assays. Histograms of mean relative luciferase activity in HEK293 cells co-transfected with a dual reporter vector (pmirGLO) containing cloned DNA sequences corresponding to the 3′ UTR of the indicated genes plus synthetic miRNA duplexes representing miR-96, miR-182, miR-183, or all three (ALL) normalized to scrambled siRNA control (CTRL). DNA sequences corresponding to the 3′ UTRs of Sox2, Notch1, Jag1, and Hes1 were inserted downstream of the pmirGLO Photinus open reading frame. For each assay, the ratio of Photinus and Renilla luciferase activity in cells co-transfected with reporter vector and synthetic miRNA was normalized to that of cells transfected with reporter vector and scrambled siRNA. Each bar represents two replicate readings from each of six transfections performed over three experiments. Error bars indicate the standard error of the normalized mean. The Wilcoxon signed rank test was used to determine statistically significant differences in relative luciferase activity compared to control. Asterisks indicate P < 0.01 (*) and P < 0.005 (**).  While there appears to be compelling evidence for wider effects on conserved miRNA targets in P18-cochlea, the hypergeometric analysis did not reveal convincing miR-183 cluster effects in the P18-cochlea microarray data. One possibility is that miR-183 cluster primary effects on gene expression observed at P5 are masked by increasing secondary effects at P18. Alternatively, the miRNA/mRNA interactions could include non-conserved and/or off-target effects of the miR-183 cluster. To address the latter two possibilities, we employed the Sylamer program to identify, through hypergeometric analysis, over-or under-represented 7-mer sequence strings (seeds) in the mouse 3′UTR transcriptome interrogated by the Affymetrix MoGene 1.0 platform 34 . Sylamer differs from the previous Targetscan-based methodology in two important ways. The first is that all gene expression data is included, regardless of probeset P-values, to generate ranked gene-lists from most upregulated to most downregulated. The second is that evolutionary conservation of predicted miRNA seed sequences across homologous 3′UTRs is not a component in the analysis. Sylamer analysis of 912 miRNA target heptamers, including all conserved 8mer, 7mer-m8, 7mer-1A sites from TargetScan 6.2, revealed significant depletion of two miR-183 cluster target heptamers (GTGCCAA, TGCCAAA) in downregulated Tg 1MDW/1MDW P5-OC 3′UTRs (Fig. 6A), complementing the TargetScan 6.2-based hypergeometric analysis (Table 5, Fisher's Exact Test P = 0.017). This transcriptome level microarray analysis validates the engineered intent of Tg 1MDW , i.e. miR-183 cluster SC misexpression to downregulate genes that are disproportionate with respect to miR-183 cluster 3′UTR-bearing seed sequences. For P18-cochlea, Sylamer analysis showed no significant (conserved, non-conserved and/or off-target) effects attributable to miR-183 cluster (Fig. 6B). Nevertheless, 49 conserved miRNA target heptamers were enriched (17) or depleted (32) (Fig. 6B). However, these miRNA heptamer seeds were distinct from those conserved miRNAs identified by TargetScan 6.2-based hypergeometric analysis (Table 5, Fisher's Exact Test P = 0.0033). As confirmation of our in-house application of Sylamer, we also analyzed the publicly available microarray data (n = 3), confirming enrichment of miR-96 wt heptamers (GTGCCAA, TGCCAAA) and depletion of miR-96 ddl heptamers (GAGCCAA, AGCCAAA) in upregulated and downregulated genes, respectively, in the diminuendo (Dmdo or Mir96 ddl/ddl ) OC at P4 2 (Fig. 6C) in agreement with the TargetScan 6.2 based hypergeometric analysis (Table 5, Fisher's Exact Test P = 0.013).

Trangenic miRNA misexpression results in profound cochlear dysfunction. Homozygous
Tg 1MDW/1MDW mice lose Preyer's reflex, an indication of hearing loss 24 . Consistent with that, we found evidence of OHC loss in the cochlear base at P18 (see Fig. 3). Moreover, phalloidin staining of cochlear whole mounts demonstrate missing OHCs in the basal OC at P21 (Fig. 7A-D) and stereocilia defects, which includes membrane fusion and bundle architecture defects at P30 (Fig. 7E,F). Immunofluorescence of whole mount cochleae at P37 labeled with primary antibodies to MYO7a confirm a base-to-apex degeneration of cochlear HCs (Fig. 8), with many in Tg 1MDW/1MDW OHCs altered from a cylindrical shape to being significantly shorter with larger diameter basal ends compared to WT OHC (Fig. 8). By P115, very few IHC or OHC remain (data not shown).
Consistent with histological analyses, average auditory brainstem response (ABR) threshold (dB SPL) vs. frequency curves demonstrate progressive hearing loss from P18 to P90 in homozygous Tg 1MDW/1MDW mice (Fig. 9A). Responses were not detected even at the highest levels for stimulus frequencies >32 kHz between P18 and P35, nor were responses detected at frequencies >22.6 kHz in P90 Tg 1MDW/1MDW mice. A two-way mixed analysis of variance of ABR threshold as a function of stimulus frequency from animals at P90 yielded a significant genotype X stimulus interaction, F(16,120) = 30.89, P < 0.001, a significant main effect of genotype, F(2,15) = 1228.48, P < 0.001 and a significant main effect of stimulus, F(8,120) = 119.72, P < 0.001. Bonferroni corrected post-hoc tests showed significant threshold differences between WT and homozygous Tg 1MDW/1MDW animals (P < 0.001), and hemizygous and homozygous Tg 1MDW/1MDW animals (P < 0.001), but not between WT  and hemizygous animals (P = 0.067). One-way ANOVAs showed significant differences among genotypes for all stimuli tested and multiple comparison tests indicated that thresholds of homozygous Tg 1MDW/1MDW mice were significantly higher than WT and hemizygous mice for all stimuli tested. Average distortion product otoacoustic emission (DPOAE) amplitude frequency curves confirm the peripheral sensory nature of hearing loss in Tg 1MDW/1MDW mice (Fig. 9B). Although DPOAEs were observed at P90 in Tg 1MDW/1MDW mice, responses were near the noise floor. A two-way mixed ANOVA of DPOAE level as a function of f2 frequency from animals at P90 yielded a significant interaction between frequency and genotype, F(64,480) = 6.1 P < 0.001, a significant main effect of frequency, F(32,480) = 23.59, P < 0.001, and a significant main effect of genotype, F(2,15) = 9.9, P < 0.05 (P = 0.002). Bonferroni corrected post-hoc tests showed significant amplitude differences between WT and homozygous Tg 1MDW/1MDW animals (p = 0.01), as well as between hemizygous and homozygous Tg 1MDW/1MDW animals (P = 0.003) but not between amplitudes of WT and hemizygotes (P = 1.0). One-way ANOVAs indicated significant differences among genotypes for f 2 frequencies from 7.3 kHz to 29.3 kHz and multiple comparison tests showed significant differences in DPOAE level between homozygous Tg 1MDW/1MDW mice and both WT and hemizygous mice, but not between WT and hemizygous mice.
Stimulus level-dependent, progressive hearing loss was also apparent as shown in a series of DP amplitude vs. stimulus level curves representing responses to an f2 frequency of 12.8 kHz (Fig. 9C). Progressively increasing DPOAE thresholds with age were observed, findings that are consistent with the observed inner ear pathology, and confirming that the hearing impairment reported here results from sensory HC degeneration. A 2-way mixed ANOVA of DPOAE level as a function of f1 level from animals at P90 yielded a significant interaction between level and genotype, F(26,169) = 35.9 P < 0.001, a significant main effect of level, F(13,169) = 178.9, P < 0.001, and a significant main effect of genotype, F(2,13) = 56.8, P < 0.001. Bonferroni corrected post-hoc tests showed significant amplitude differences between WT and homozygous Tg 1MDW/1MDW animals (P < 0.001), as well as between hemizygous and homozygous Tg 1MDW/1MDW animals (P < 0.001), but not between amplitudes of WT and hemizygotes (P = 1.0). One-way ANOVAs showed significant differences among genotypes at levels at and above 30 dB SPL and multiple comparison tests indicated significant differences between homozygous Tg 1MDW/1MDW and WT mice at levels >25 dB SPL and between homozygous Tg 1MDW/1MDW and hemizygous mice at levels >35 dB SPL, although, responses for homozygous Tg 1MDW/1MDW mice were at noise floor levels below 70 dB SPL.
Both ABR and DPOAE measures correlate with the degree of histological HC loss in Tg 1MDW/1MDW mice (Figs. 6, 7). Overall, sensory HC degeneration in Tg 1MDW/1MDW mice suggests a significant potency of these three miRNAs in effecting tissue homeostasis through GFAP promoter-driven miRNA-183 cluster misexpression and represents a novel biological reagent useful to identify molecular pathways and mRNAs targeted by the miR-183 cluster.

Discussion
Transgenic Tg 1MDW/1MDW mice were developed to misexpress miR-183 cluster in OC SCs to further explore the role of these HC-specific miRNAs on OC cell differentiation. In WT OC, the cell-specific mutual exclusion of miR-183 cluster expression from their respective mRNA targets ensures that HCs enforce repression of genes both temporally and spatially, relative to those same genes in SCs. The mutual exclusion hypothesis of miRNA function 16 predicts that in Tg 1MDW/1MDW , miR-183 cluster misexpression, driven by the GFAP promoter in adjacent and lineage-related SCs (compare Fig. 1A,B) should repress evolutionarily conserved target genes that are simultaneously essential to SC function and incompatible for attaining and/or maintaining the HC fate. The majority of downregulated genes observed by microarray (Table 2) at P5 are glial cell specific, supporting a significant influence of miR-183 cluster gene regulation on glial cell misexpression, most likely from Schwann cells that underlie the greater epithelial ridge (GER) and inner sulcus in the P5 OC microdissected tissue. At P6, possible effects on OC cell differentiation in Tg 1MDW/1MDW were revealed by detecting a modest increase in medially placed IHCs ( Fig. 2A-C), and by the observation of SC types (Deiters' cells) that express the HC marker MYO6 (Fig. 3A,B). Whether these cells are the result of SC transdifferentiation and/or are generated through perturbations of cell-cell signaling pathways known to effect such changes, such as Wnt and Notch signaling 35 , is the focus of our ongoing studies. Previous studies with this cluster have shown them to be potent modulators of HC fates in zebrafish in vivo 11 . These extra IHCs phenocopy single-gene hypomorphic and null mutations in genes that specify SC identity and that are predicted targets of miR-183 cluster miRNAs (Fig. 2D) [25][26][27][28][29][36][37][38] . Indeed, our dual luciferase assays validated Sox2, Notch1 and Hes1 3′UTRs as targets for post-transcriptional repression by miR-183 cluster members (Fig. 4). Cochlear HCs and adjacent SCs of the mammalian OC arise from post-mitotic prosensory precursors within specialized otic epithelium 20 . Notch1, Jagged1 (Jag1, a Notch ligand) and a downstream transcription factor (TF), Sox2, define this precursor pool through the Notch-mediated process of lateral induction 39 . Notch signaling also governs the process of lateral inhibition, which results in sharply contrasting cell types in the OC. In this process, presumptive HCs upregulate the TF Atoh1, notch ligands Delta1 (Dll1) and Jagged 2 (Jag2), and the miR-183 cluster. Dll1 and Jag2 increase notch1 activation in adjacent SC cells. This activation releases the notch intracellular fragment, N ICD , inducing genes that inhibit HC fates and promote SC fates 40 , including expression of GFAP, a glial-specific intermediate filament protein 41 . Interestingly, Sox2 persists in adult OC SCs, suggesting a maintenance of precursor identity 42 . Since Notch signaling is regulated by miRNAs in Drosophila 43 , one testable hypothesis using Tg 1MDW/1MDW is to ask whether SC miR-183 cluster expression affects Notch-mediated cell fate specification and/or homeostasis as the mechanism for how targeted misexpression of the HC specific miR-183 cluster may increase the propensity of SCs to transdifferentiate into HCs, an emerging paradigm for treating hearing loss [44][45][46] .
Rapid age-related demise of HCs, observed in this model both histologically (Figs 7, 8) and physiologically (Fig. 9), suggests a specific potency to dysregulation of miR-183 cluster target genes on postnatal OC homeostasis and function. While it can't be ruled out that integration of Tg 1MDW may have disrupted an endogenous gene critical for hair cell survival, comparing Tg 1MDW/1MDW whole transcriptome effects to those from previously published miR-183 cluster mouse loss-of-function mutants (Fig. 6, Table 5) revealed reciprocal effects on miR-183 cluster target sites. These reciprocal effects are consistent with a reduction of miR-183 cluster miRNAs in sensory cells 2,33 versus a gain of miR-183 cluster miRNAs in SCs (this study). Despite these fundamental miRNA target-specific transcriptome differences, sensory cell demise predominates in all three mouse models. Interestingly, comparing the 19,358 interrogated genes common to both P5 OC Tg 1MDW/1MDW versus P3 OC Mir96 ddl/ddl microarray studies, the overlap of Mir96 ddl/ddl changed genes (n = 97) with Tg 1MDW/1MDW changed (179) genes (uncorrected P < 0.05, fold change cutoff ± 1.5) were statistically not independent (Odds Ratio 7.27, Fisher Exact test P = 0.00028) with all 6 common genes changed in the same direction (5 downregulated : Ncmap, Pmp22, Prx, Mag, Fa2h, 1 upregulated: Stfa1). So, while there is evidence for reciprocal effects on miR-183 cluster target sites, these models do exhibit common gene expression changes, as well. Indeed, the common HC phenotypes in regards to stereocilia defects, ABR threshold elevation and HC death, while more severe in Mir96 +/ddl heterozygotes, are similar and suggest a narrow range of tolerance for modulations in miR-183 cluster miRNA levels, and therefore the genes they regulate.  This mammalian miRNA misexpression model demonstrates the potency of small non-protein coding miR-NAs and should be useful in genomic/transcriptomic/proteomic studies to identify primary miR-183 cluster target genes and regulatory, structural and/or metabolic pathways affected by their dysregulation. The elucidation of miRNA-regulated pathways affected in SCs may provide novel avenues for future therapeutic interventions in treating some forms of hearing loss, be it by identifying molecular genetic pathways critical to HC homeostasis or by informing SC transdifferentiation research.

Animals. Animal studies were approved by both Creighton University and Boys Town National Research
Hospital Institutional Animal Care and Use Committees and were consistent with the National Research Council Guide for the Care and Use of Laboratory Animals (2011). FVB/NCrl mice were purchased from Charles River Laboratories.

Generation of transgenic mice and genotyping.
A 453 bp DNA fragment encompassing the pre-miRNA sequences of both miR183 and miR96 were PCR amplified from total mouse DNA using the following primers: 5′-CAGTCCCGGGTGCAGGCTGGAGAGTGTGAC-3′ and 5′-GATCGATATCCCTCAGGCAGTGAAAGGTGA TC-3′. A 586 bp region including pre-miR-182 was PCR amplified using primers 5′-CATGGATATCGGGCTT GAGGAGGTTTTACAC-3′ and 5′-GTACGCGGCCGCGATCGCATAGACCAGAAGACAC-3". The miR-183/miR-96 PCR product was directionally cloned into the XmaI-EcoRV sites within the polycloning region of pIRES-hrGFPII (Stratagene) to create p183-X-E. The miR-182 PCR product was subsequently cloned into the EcoRV-NotI sites of p183-X-E to create p182-10. The cloned sequences for both plasmids were verified by DNA sequencing. HEK293 cells transfected with p182-10 verified expression and processing of significantly elevated levels of mature miR-183 cluster miRNAs using commercial miRNA PCR assays (Ambion, data not shown). To generate the miR-183 cluster transgene, a 1045 bp PCR product having Mir183, Mir96 and Mir182 was amplified from p182-10 using the following PCR primers: 5′-CAG TAG ATC TTG CAG GCT GGA GAG TGT GAC-3′ and 5′-CTG AAG ATC TGA TCG CAT AGA CCA GAA GAC AC-3. This PCR product was digested with BglII and replaced the nls-LacZ BamHI fragment of pGFA2-nlac (a gift from M. Brenner, UAB) to create pGfa2-miRs-183-96-182. A linear 3.8 kb NspI fragment from pGfa2-miRs-183-96-182 (genbank ID: JX912274) was purified and used for pronuclear microinjection into FVB/NCrl embryos. Three founders (Tg 1MDW , Tg 2MDW , Tg 3MDW ) were identified by PCR amplification of a 512 bp DNA product using the following primers: 5′-TTG GCA ATG GTA GAA CTC ACA C-3′ and 5′ATC TGC TCC TGC TTT TGC TG-3. Independent transgenic lines of the founders were established by backcrossing to FVB/NCrl and maintained by filial mating. Real-time quantitative PCR was used to genotype animals for genetic, molecular and histologic studies described below. Briefly, 5′ nuclease assay primer/probe sets for Tg(GFAP-Mir183,Mir96,Mir182) and endogenous Atoh1 gene were simultaneously amplified in PCR reactions with Platinum ® Quantitative PCR SuperMix-UDG w/ROX (Invitrogen) using DNA purified from tail biopsies. Table 6 lists the primer sequences used for quantitative Tg genotyping. PCR was run on a StepOnePlus ™ PCR system (Applied Biosystems ® ). Tg genotypes were determined using the ΔΔCt method using Atoh1 as the normalization control. The hemizygote and homozygote quantitative PCR distributions within each transgenic line were statistically distinct (Fig. 1).

Histologic analysis.
Whole-mount in situ hybridization (ISH) using LNA-DIG labeled probes against miR-182 was performed as previously described 7 . Cy5 fluorescence of miR-182 used sheep anti-DIG-POD Fab antibodies (Roche) with a tyramine signal amplification kit (TSA Plus, Perkin Elmer). For fluorescent labeling of F-actin, mouse cochleae were harvested, microdissected, fixed in 4% PFA overnight, dehydrated-rehydrated through an ethanol series, blocked with normal goat serum and stained overnight with TRITC-Phalloidin (Sigma, 0.5ug/ml), rinsed in PBS, mounted on slides with glycerol and imaged using a Zeiss LSM 510 META LNO confocal microscope. For electron microscopy, ears were fixed overnight in 2.5% glutaraldehyde, postfixed in 1% OsO4 in cacodylate buffer, dehydrated, sputter coated and mounted for scanning electron microscopy (SEM).
For immunohistochemistry, tissue pieces of the OC (apex, middle and base) were block/permeabilized with 5% Normal Goat Serum (NGS)/0.1% Tween20 at room temperature for 2-3 hours, incubated with polyclonal rabbit MYO6 or MYO7a primary antibodies (Proteus Biosciences, #25-6791-MYO6, #25-6790-MYO7a) for 48 hours in blocking buffer and washed three times with PBS, labeled with either Alexa 488-or Alexa-568-conjugated anti-rabbit secondary antibodies (1:500) (Invitrogen), washed with PBS, coverslipped using Prolong anti-fade (Invitrogen), and analyzed using either a Zeiss LMS 510 or LMS 800 confocal microscope.   Figure S1) or snoRNA-135 as the normalization control.    Cochlear function. Auditory brainstem responses (ABRs) and distortion product otoacoustic emissions (DPOAEs) were recorded in WT, Tg 1MDW hemizygous and Tg 1MDW/1MDW homozygous mice anesthetized with a ketamine-xylazine mixture (100 mg/kg ketamine, 15 mg/kg xylazine IP) and supplemented with 25-50% of the initial dose as needed throughout the recording session to maintain a stable, quiet recording environment. Body temperature was controlled using a thermostatically regulated heating blanket and thermal probe, and body temperature was maintained at approximately 37.5 °C (Harvard Apparatus). Heart rate was monitored throughout the procedure and fluids were replaced as needed. All recordings were conducted in an electrically shielded, double-walled, sound-attenuating chamber (Industrial Acoustics Corp).

ABR Procedures.
ABRs were used to assess the integrity of the cochlea and auditory brainstem non-invasively as detailed previously 50,51 . Stimuli consisted of 3 ms pure tone bursts (1 ms raised cosine on/off ramps and 1 ms plateau) or 64μs clicks. Both tone bursts and clicks were digitally generated (125 kHz clock rate) and delivered free-field through a high impedance piezoelectric tweeter (Radio Shack). For stimuli above 32 kHz, an electrostatic speaker (ES1, TDT) was substituted. Sound sources were placed ~10 cm from the cranial vertex. Stimulus levels were calibrated in decibels sound pressure level (dB SPL) with a 1/8-inch Brüel and Kjaer microphone (Model 4138). Platinum needle electrodes (Grass Instruments) were positioned subdermally at the vertex (active, non-inverting), the infra-auricular region (reference, inverting), and the neck region (ground). Scalp voltage potentials were amplified 100,000X, band-pass filtered between 0.03 and 10 kHz (Grass Model P511K), and digitized (Tucker-Davis Technologies, TDT) at a 20 kHz sampling rate over 15 milliseconds. A total of 200 trials were averaged for each response. Waveforms were stored digitally for off-line analyses and custom software was used for data acquisition and subsequent data analyses. ABR thresholds were determined for clicks and for tone bursts in half octave steps ranging from 64 to 2.0 kHz. Stimulus levels were decremented from 90 dB SPL to below threshold in 10 dB steps. Threshold values were subsequently refined using a bracketing strategy in which level was adjusted in 5 dB steps relative to threshold values previously determined in 10 dB step decrements. Threshold was defined as the smallest stimulus that generated an unambiguous, replicable response.
Briefly, 2 phase-locked pure tone stimuli (f1, f2) were generated by 24-bit D/A converters (Lynx22 soundcard) and conveyed through separate earphones. The primary tone frequencies were presented such that f2/f1 = 1.25, and f2 intensity was 10 dB lower than f1. The earphone outputs, along with a low-distortion probe microphone (Etymotic Research, ER-10B+) were sealed within the external ear canal, forming a closed acoustic system.  Acoustic emissions recorded by the probe microphone were amplified 40 dB and sampled with a 24-bit A/D converter (Lynx Studio Technology, L22) at 48 kHz for tone frequencies below 5 kHz or at 192 kHz at tone frequencies above 5 kHz. FFT analyses were used to compute level/phase of component DPOAEs and their corresponding noise floors. Physiological results were analyzed using a two-way mixed analysis of variance (IBM SPSS version 22), with genotype (WT, Tg 1MDW hemizygous and Tg 1MDW/1MDW homozygous groups) as the between-subject variable. For ABRs, stimulus type (click and frequencies from 2 to 22.6 kHz) was the repeated measure, for DPOAE frequency sweeps, f2 frequency (4 to 64 kHz) was the repeated measure, and for DPOAE input-output curves, f1 level (10 to 75 dB SPL) was the repeated measure. Bonferroni adjustments were made for multiple comparisons and data were analyzed further using one-way analysis of variance (ANOVA) and Dunnett C tests were used for multiple comparisons. Differences between means were considered statistically significant when P < 0.05.