Introduction

MicroRNAs are short (~20–24nt) endogenous non-coding RNAs that posttranscriptionally repress the expression of protein-coding genes. The effects mediated by miRNAs on inner ear morphogenesis, neurosensory cell identity, function and homeostasis indicate that gene regulation through miRNAs are critical to the biology of the inner ear1,2,3,4,5,6. The polycistronic cluster of Mir96, Mir182 and Mir183 genes are abundantly expressed in afferent cochlear and vestibular neurons and their peripheral innervating targets: auditory and vestibular hair cells (HCs)7,8,9. From an evolutionary viewpoint, the miR-183 cluster of miRNA genes are syntenic, highly conserved and co-expressed in neurosensory organs of animals representing several taxonomic phyla, suggesting that the control of gene expression through by this miRNA cluster is highly coordinated and under extraordinary selective pressure10.

The spatiotemporal expression pattern of Mir96, Mir182 and Mir183 in the developing vertebrate inner ear and the effects induced by modulating levels of these miRNAs on HC fate determination in zebrafish and chicken argue that these miRNAs collectively function, to some degree, in the transition from inner ear prosensory cells towards a HC fate8,9,11,12. Also, they have significant roles in HC functional maturation and homeostasis2,5,13,14. Genome-wide studies demonstrate that miRNAs can fine-tune mRNA levels in the context of miRNA/target mRNA co-expression and can also act as developmental on-off switches, as in the context of mutually exclusive miRNA/target mRNA expression15,16,17,18,19. For those effects attributable to the miR-183 cluster in the vertebrate inner ear, these functional distinctions in miRNA regulation might be segregated temporally: in a switch-like manner 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 signaling20. By E15.5, miR-183 cluster expression distinguishes differentiated cochlear HC, which suggests a role for these miRNAs in this cell type transition9. 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 strategies6. To test this hypothesis, we engineered Tg(GFAP-Mir183,Mir96,Mir182) (Tg1MDW) mice to drive ectopic miR-183 cluster expression using the core human promoter of the glial fibrillary acidic protein (GFAP). Overall, Tg1MDW 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.

Table 1 Tg-(GFAP-Mir183,Mir96,Mir182) line information.

Results

FVB/NClr-Tg(GFAP-Mir183,Mir96,Mir182)1MDW mice

SCs in the postnatal inner ear organ of Corti (OC) express endogenous GFAP and human GFAP-promoter driven GFP and LacZ reporters21,22. To direct misexpression of miR-96, miR-182 and miR-183 in the SCs of the inner ear, we modified an established GFAP promoter-driven reporter construct (pGFA-nlac, Michael Brenner, UAB) by substituting the nLacZ gene with the miR-183 cluster coding sequences23. Of 425 implanted embryos, 3 of the 16 live born pups were identified as founders and three independent inbred lines were established in FVB/NClr: Tg(GFAP-Mir183,Mir96,Mir182)1MDW, Tg(GFAP-Mir183,Mir96,Mir182)2MDW, Tg(GFAP-Mir183,Mir96,Mir182)3MDW (Tg1MDW, Tg2MDW, Tg3MDW). All lines transmit at single-gene Mendelian ratios and map to unique cytogenetic loci (Table 1). At a gross phenotypic level, adult homozygous Tg1MDW/1MDW mice lose Preyer’s reflex, an indication of hearing loss24. Both homozygous and hemizygous mice from Lines Tg1MDW and Tg2MDW 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 Tg3MDW 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 (Tg1MDW >>> Tg2MDW > Tg3MDW, 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 Tg1MDW cerebellum only, a common cell type expressing GFAP-core promoter transgenes23 (Supplementary Figure 1B–D).

MicroRNA 183 cluster is misexpressed in Tg1MDW/1MDW SCs

Loss of Preyer’s reflex in Tg1MDW/1MDW adults led us to concentrate our inner ear characterizations to this genotype. We confirmed transgenic SC miR-183 cluster expression directly by dual whole mount ISH/IHC using LNA-DIG labeled probes against miR-182 and an antibody against MYO6 (Fig. 1A–D). Compared to the WT staining of miR-182 (Fig. 1A) that was restricted to inner and outer HCs (Fig. 1C), nuclear miR-182 labeling was found in all Tg1MDW/1MDW SCs (Fig. 1B, arrowheads), including Deiters, inner pillar, outer pillar, inner phalangeal and inner border cells. Spiral limbus and Schwann cell nuclei also were stained with LNA-miR-182, similar to other GFAP promoter driven transgenes (Fig. 1B, arrows, Smeti et al., 2011). Unfortunately, we were unable to identify clear differences in miR-183 cluster expression in WT versus Tg1MDW/1MDW at earlier ages (i.e. P5 and P10, data not shown). The Tg1MDW construct lacked a separate reporter gene (e.g. LacZ, GFP) for this purpose. However, further evidence of transgenic SC miR-183 cluster expression was obtained through quantitative RT-PCR, which showed 2.9-, 2.7- and 2.2-fold higher levels of miR-182, miR-96 and miR-183, respectively, in Tg1MDW/1MDW versus WT P18-cochlear total RNA (Fig. 1E). Taken together, these results confirm ectopic miR-183 cluster expression in OC SCs, spiral limbus cells and the Schwann cells that ensheathe neuronal processes projecting to and from the OC.

Figure 1
figure 1

Supporting cell (SC) expression of miR-182 in Tg1MDW/1MDW mice at P18. (AD) Dual ISH/IHC fluorescence imaging of whole mount organ of Corti (OC) from WT (A,C) and Tg1MDW/1MDW (B,D) P18 littermates. Cy5 fluorescence of tyramide enhanced miR-182 ISH labeling shows nuclear/cytoplasmic staining in hair cells (HCs A,B) and definitive Tg1MDW/1MDW OC SC nuclear staining, including Deiters’, inner pillar, outer pillar, inner phalangeal and inner border cells (B, arrowheads). SC labeling with miR-182 was found in spiral limbus cells and myelinated Schwann cells of spiral ganglion neurons (B, arrows). Alexa 546 immunofluorescence of MYO6 positive HCs (C,D). (E) Quantitative increase in miR-183 cluster in P18 Tg1MDW/1MDW total cochlear RNA by RT-PCR. Using ABI Taqman assays, quantitative RT-PCR was performed in Tg1MDW/1MDW versus WT littermates (n = 3). The miR-183, miR-96 and miR-182 levels were normalized to Sno135. The results quantitate statistically significant (ΔCT values, 2 sample t-test, P < 0.001) increases in miR-182 (2.9 fold), miR-96 (2.7 fold) and miR-183 (2.2 fold) in Tg1MDW/1MDW cochlea at P18. These data are consistent with a transgenic misexpression of these neurosensory miRNAs via the core human GFAP promoter and validate microarray data. Scale bar: 20 μM (AD).

Tg1MDW/1MDW OC show increased cochlear IHCs and MYO6 positive Deiter’s Cells

Even though we could not detect Tg1MDW/1MDW misexpression at early postnatal ages by ISH, a clear histomorphological change in Tg1MDW/1MDW OC were observed around P6 (Fig. 2A,B). A modest increase in apparent IHCs was found (Fig. 2B, arrows). Because cells in the OC undergo significant postmitotic functional maturation and programmed apoptosis occurs medially in the greater epithelial ridge, we chose P18 to quantify the expression of additional IHCs. When compared with WT littermates, Tg1MDW/1MDW had an average of 7.2 extra IHC/mm of apical OC versus 0.97 extra IHC/mm in WT (Fig. 2C,D, n = 3 ears, T-test P = 0.038). The presence of additional HCs in this mouse model of SC miR-183 cluster misexpression is consistent with previous studies in both zebrafish and chicken where overexpression of miR-183 cluster miRNAs increased inner ear HC numbers11,12. Interestingly, this extra IHC patterning in Tg1MDW/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 mice26,27. For Sox2, extra IHCs were similarly extrapolated from counts taken from the entire length of late embryonic (E18) Sox2+/EGFP reporter knockout (KO) and SoxEGFP/LP hypomorphic cochleae28. For Hes1, counts were from 1 mm lengths starting from the basal end of E18 cochleae of KO Hes1+/− and Hes1−/− mice25. 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.

Figure 2
figure 2

Excess IHCs are a consistent phenotype of prosensory/SC gene mutations. (A,B) Whole mount miR-182 staining of HCs in the cochlear apex of P6 WT (A) and Tg1MDW/1MDW (B) mice. Arrows point to additional miR-182 positive, medially placed IHCs. (C) Phalloidin-stained Tg1MDW/1MDW OC at P18 with 4 contiguous, medially placed extra IHCs (arrow heads). (D) Mean extra IHCs per 1 mm distance was plotted and compared with other published mouse mutants. Mutations in genes that are predicted targets of the miR-183 cluster phenocopy Tg1MDW/1MDW-mediated increases in IHCs. Phenotypic correlation of extra IHCs in SC-specific single-gene mutants are consistent with negative regulation of these predicted miR-183 cluster target genes. Scale bar: 200 μM (A,B); 20 μM (C).

Figure 3
figure 3

Deiter’s cell expression of MYO6 in P18 Tg1MDW/1MDW OC, an early marker of differentiated HCs. (A) MYO6 staining in OC whole mount showing definitive label in Deiter’s cells (black arrowheads). (B) miR-182 co-staining. MYO6 is a marker of HC differentiation, suggesting that miR-183 cluster misexpression and its cytoplasmic localization are positive effectors of HC identity. Scale bar: 20 μM (A,B).

miR-183 cluster mediated translational repression of select target genes

To determine whether miR-183, miR-182, or miR-96 directly regulate select SC genes important to HC/SC differentiation (i.e. Jag1, Sox2, Hes1, Notch1) dual luciferase assays were performed. The respective 3′ UTRs were PCR amplified and cloned into the pmirGLO vector immediately downstream of the Photinus luciferase coding sequence. Each were co-transfected in HEK293 cultures with synthetic RNA duplexes representing miR-96, miR-182, or miR-183 alone or in combination. Scrambled miRNA duplexes were co-transfected with the reporter vector as a control. 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 target9. 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 repression30. 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 differentiation28, while Notch1-regulated Hes1 gene expression also antagonizes the ability of Atoh1 to promote hair cell differentiation31. Given that miR-183 cluster expression requires Atoh1-mediated HC specification9, 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.

Figure 4
figure 4

Validation of Sox2, Notch1, 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 (**).

Microarray analysis reveals miR-183 cluster-specific changes in P5-OC transcriptome

The mutual exclusion hypothesis of miRNA function16 predicts that miR-183 cluster misexpression in adjacent and lineage-related SCs should repress evolutionarily conserved target genes essential to SC identity/function. To test for this type of aggregate effect, we performed whole-genome transcription profiling using Affymetrix microarrays (MoGene 1.0 microarrays) to evaluate mRNA expression in Tg1MDW/1MDW versus WT in the P5-OC (n = 5) and P18-cochlea (n = 3). Two-way ANOVA testing found significant differences (FDR-adjusted P < 0.05, fold change cutoff ± 2) in Tg1MDW/1MDW versus WT littermates at either P5 (Table 2, Fig. 5A) or P18 (Table 3, Fig. 5B). Irrespective of age, the most upregulated Affymetrix probesets were genetic elements of Tg1MDW: Mir183, Mir96, Mir182 and Prm1 (Fig. 5A,B, boxed probesets), with fold changes + 5.67 (P18-Prm1) to + 16.97 (P18-Mir182). In the P5-OC, the majority of changed genes (27/32) were downregulated (Fig. 5A, Table 2), whereas in P18-cochlea, the majority were upregulated (37/45), not counting Prm1 and miR-183 cluster miRNAs (Fig. 5B, Table 3).

Table 2 List of 32 significantly changed Tg1MDW/1MDW Affymetix probesets in the P5-OC. From a total of 28,188 annotated probesets using the Limma Package and MicroArraysRUS Gui interface to R and Bioconductor microarray analysis programs.
Figure 5
figure 5

Clustered heatmaps of significantly changed genes in Tg1MDW/1MDW Affymetrix microarray analysis. (A) OC-P5 heatmap showing normalized expression for 32 significantly changed genes found by two-way ANOVA (Age X Genotype, FDR-adjusted P < 0.05, fold change cutoff ± 2). (B) Cochlear-P18 heatmap of normalized expression for 52 probesets (45 genes) significantly changed in Tg1MDW/1MDW (FDR-adjusted P < 0.05, fold change cutoff ±2). Irrespective of age, the most upregulated Affymetrix probesets were genetic elements of Tg1MDW: Mir183, Mir96, Mir182 and Prm1 (boxed probesets).

Table 3 List of 52 significantly changed Tg1MDW/1MDW changed Affymetix probesets in the P18-cochlea. From a total of 28,188 annotated probesets using the Limma Package and MicroArraysRUS Gui interface to R and Bioconductor microarray analysis programs.

To provide a biologic context to these expression changes, we performed gene ontology (GO) analysis using the DAVID bioinformatics resources32. 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 miRNAs and 62,793 3′UTR seed sites (8mer, 7mer-m8, 7mer-1A, 3comp, TargetScan 6.229). 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 Tg1MDW/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 Tg1MDW/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.

Table 4 Overrepresented Gene Ontology (G) terms for upregulated and downregulated genes in P5-OC and P18-cochlea. FDR < 0.05, fold-change ± 1.5. Gene totals (Up|Down): P5-OC (27|93); P18-cochlea (15|86).
Table 5 Hypergeometric analysis of genome-wide microarray data demonstrate miR-183 cluster specific enrichment/depletion of evolutionary miRNA/3′UTR seed targets.

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 (miR96ddl/ddl)2 and from 5wk-retina miR-183 cluster KO mouse (miR-183CGT/GT)33. In the case of the miR96ddl/ddl microarray, miR-96 seed sites were enriched (Odds Ratio 2.10, FDR-adjusted P < 0.05) in miR96ddl/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 mutation2. In miR-183CGT/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-183CGT/GT upregulated genes (uncorrected P < 0.05, fold change >+ 1.05), but also the reciprocal depletion of miR-96 and miR-182 targets (Odds Ratio 0.52, 0.58, respectively, FDR-adjusted P < 0.05, Table 5) in miR-183CGT/GT downregulated genes (uncorrected P < 0.05, fold change <−1.05). Taken together, the hypergeometric analysis of Tg1MDW/1MDW demonstrates that at P5 in the OC, miR-183 cluster target genes, defined by both miRNA and 3′UTR seed site evolutionary conservation, are reduced concomitantly with transgenic overexpression of miR-183, mir-96 and miR-182.

Table 6 Tg-(GFAP-Mir183,Mir96,Mir182) genotyping primers and probes.

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 platform34. 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 Tg1MDW/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 Tg1MDW, 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-96wt heptamers (GTGCCAA, TGCCAAA) and depletion of miR-96ddl heptamers (GAGCCAA, AGCCAAA) in upregulated and downregulated genes, respectively, in the diminuendo (Dmdo or Mir96ddl/ddl) OC at P42 (Fig. 6C) in agreement with the TargetScan 6.2 based hypergeometric analysis (Table 5, Fisher’s Exact Test P = 0.013).

Figure 6
figure 6

Sylamer analysis of microarray data showing enrichment and depletion for 911 miRNA heptamer seed sequences in 3′UTRs. (A) Tg1MDW/1MDW P5-OC demonstrates significant depletion of 2 miR-183 cluster heptamers (GTGCCAA, TGCCAAA), validating this miRNA misexpression model. The x axis is the number of all microarray genes sorted from most upregulated (left) to most downregulated (right). The y axis shows the hypergeometric significance for enrichment or depletion of heptamers in 3′UTRs at leading parts of the gene list. Positive values indicate enrichment [-log10(P)] and negative values depletion [log10(P)]. Horizontal lines represent 0.05 Bonferroni P-value threshold. (B) Tg1MDW/1MDW P18-cochlea microarray fails to demonstrate miR-183 cluster effects, but does show significant imbalances in 3′UTR heptamer representation for dozens of miRNA complementary sequences. (C) Analysis of available primary Mir96ddl/ddl P3-OC microarray data (n = 3) confirms enrichment of miR-96wt (GTGCCAA, TGCCAAA) and depletion of miR-96ddl heptamers2. All highlighted plots are miR-183 cluster heptamers and include two (2) predicted to complement Mir96ddl mutant miR-96. Sylamer program reference34.

Trangenic miRNA misexpression results in profound cochlear dysfunction

Homozygous Tg1MDW/1MDW mice lose Preyer’s reflex, an indication of hearing loss24. 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 Tg1MDW/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).

Figure 7
figure 7

Tg1MDW/1MDW OHC loss and stereocilia defects. (AD) Confocal microscopy of MTRITC-Phalloidin-stained P21 WT and Tg1MDW/1MDW cochlear tissue whole mounts. The images are brightest point projections of 7.6 micron Z-plane images parallel to the reticular lamina. (A,B) Apical OC shows comparable OHC/SC F-actin cytoarchitecture in WT versus Tg1MDW/1MDW at P21. (C,D) In the basal OC, F-actin cytoarchitecture is disturbed due to missing OHCs in Tg1MDW/1MDW at P21. (E,F) SEM images of Tg1MDW/1MDW OC reticular lamina at P30 showing stereocilia defects in both OHC and IHCs, including absence of stereocilia and stereocilia fusion (arrows). Scale bar; 20 μM (AD), 10 μM (E), 5 μM (F).

Figure 8
figure 8

Gradients of HC loss and HC dismorphology in Tg1MDW/1MDW at P37. Immunofluorescence of whole mount cochleae labeled with primary antibodies to MYO7a in P37 WT and Tg1MDW/1MDW mice. The images are brightest point projections from Z series stacks parallel to the reticular lamina. Fields were imaged from comparable locations along the cochlear duct (A,B Apex, C,D Middle, E,F Base). Note the base to apex decrease in OHC loss with the concomitant increase in shortened and rounded OHCs. The presence of extra medially located IHCs in the Tg1MDW/1MDW apex are also noted (B, white arrows). Scale bar; 50 μM.

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 Tg1MDW/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 Tg1MDW/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 Tg1MDW/1MDW animals (P < 0.001), and hemizygous and homozygous Tg1MDW/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 Tg1MDW/1MDW mice were significantly higher than WT and hemizygous mice for all stimuli tested.

Figure 9
figure 9

Hearing loss in Tg1MDW/1MDW mice. (A) ABR threshold vs. frequency curves from P18 to P90. High-frequency (>16 kHz) hearing loss was observed in replicate P18 aged Tg1MDW/1MDW homozygotes. At P22, high-frequency hearing loss of replicate homozygotes extended to frequencies above 8 kHz and by P35 hearing loss was observed across all test frequencies. By P90, the lowest thresholds observed in P90 Tg1MDW/1MDW homozygotes were in the 90 dB SPL range, and responses were not detected at the highest level of output for stimulus frequencies >22.6 kHz, nor in homozygotes from P18 to P35 at frequencies >32 kHz. (B) Average distortion product otoacoustic emission (DPOAE) amplitude vs. f2 frequency curves confirm the sensory nature of hearing loss in homozygotes. DPOAEs observed in Tg1MDW/1MDW homozygous mice were near the noise floor. (C) Average DPOAE input/output curves at an f2 frequency of 12.8 kHz indicate progressive DPOAE threshold shifts during the first 90 days.

Average distortion product otoacoustic emission (DPOAE) amplitude frequency curves confirm the peripheral sensory nature of hearing loss in Tg1MDW/1MDW mice (Fig. 9B). Although DPOAEs were observed at P90 in Tg1MDW/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 Tg1MDW/1MDW animals (p = 0.01), as well as between hemizygous and homozygous Tg1MDW/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 f2 frequencies from 7.3 kHz to 29.3 kHz and multiple comparison tests showed significant differences in DPOAE level between homozygous Tg1MDW/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 Tg1MDW/1MDW animals (P < 0.001), as well as between hemizygous and homozygous Tg1MDW/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 Tg1MDW/1MDW and WT mice at levels >25 dB SPL and between homozygous Tg1MDW/1MDW and hemizygous mice at levels >35 dB SPL, although, responses for homozygous Tg1MDW/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 Tg1MDW/1MDW mice (Figs. 6, 7). Overall, sensory HC degeneration in Tg1MDW/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 Tg1MDW/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 function16 predicts that in Tg1MDW/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 Tg1MDW/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 signaling35, 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 vivo11. 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 epithelium20. Notch1, Jagged1 (Jag1, a Notch ligand) and a downstream transcription factor (TF), Sox2, define this precursor pool through the Notch-mediated process of lateral induction39. 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, NICD, inducing genes that inhibit HC fates and promote SC fates40, including expression of GFAP, a glial-specific intermediate filament protein41. Interestingly, Sox2 persists in adult OC SCs, suggesting a maintenance of precursor identity42. Since Notch signaling is regulated by miRNAs in Drosophila43, one testable hypothesis using Tg1MDW/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 loss44,45,46.

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 Tg1MDW may have disrupted an endogenous gene critical for hair cell survival, comparing Tg1MDW/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 cells2,33 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 Tg1MDW/1MDW versus P3 OC Mir96ddl/ddl microarray studies, the overlap of Mir96ddl/ddl changed genes (n = 97) with Tg1MDW/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 miRNAs 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.

Methods

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′-GATCGATATCCCTCAGGCAGTGAAAGGTGATC-3′. A 586 bp region including pre-miR-182 was PCR amplified using primers 5′-CATGGATATCGGGCTTGAGGAGGTTTTACAC-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 (Tg1MDW, Tg2MDW, Tg3MDW) 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 described7. 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.

Real-time quantitative RT-PCR

Relative expression levels of mature miR-96, miR-82, miR-183 were assayed using ABI Taqman assays according to the ΔΔCt method using snoRNA-202 (Supplementary Figure S1) or snoRNA-135 as the normalization control.

Dual luciferase assays

DNA corresponding to the 3′ UTRs of Sox2, Notch1, Jag1, and Hes1 mRNA was amplified from mouse genomic DNA by PCR. Products were purified and inserted into the region corresponding to the 3′ UTR of the Photinus luciferase reporter gene in pmirGLO (Promega) using the In-Fusion HD Cloning System (Clontech). pmirGLO also contains a separate Renilla luciferase report gene. HEK293 cells (~2 × 105 cell/well; 24-well plate) were co-transfected using Lipofectamine 2000 (Invitrogen) with 200 ng reporter vector and 20 pmol synthetic RNA duplex representing scrambled control siRNA (Integrated DNA Technologies), miR-96, miR-182, or miR-183, or with 30 pmol combined miRNAs (10 pmol each). Cells were cultured post-transfection for 24 h and harvested to perform dual-luciferase assays using the Dual-Glo Luciferase Assay System (Promega) on a Modulus Microplate Luminometer with dual injectors (Turner Biosystems). Two replicate readings from six transfections over the course of three experiments were performed. The ratio of Photinus and Renilla luciferase activity for each reporter vector co-transfected with miRNA was normalized to that reporter vector co-transfected with scrambled control siRNA. The Wilcoxon signed-rank test was used to determine statistically significant differences in relative luciferase activity between miRNAs and control siRNA.

MicroArray Study

Ear tissues were collected and stored in RNAlater at −80 °C. Total RNA from WT and Tg1MDW/1MDW microdissected OC at P5 (n = 5) and cochleae at P18 (n = 3) were purified using the RNAeasy Mini Kit (Qiagen) following rotor-stator homogenization. The integrity and quality of each RNA preparation were determined using RNA pico chips run on a Model 2100 Agilent BioAnalyzer and quantitated using a NanoDrop 2000 spectrophotometer. RNA samples were sent to the University of Nebraska Medical Center (UNMC) microarray core facility for labeling and microarray hybridization to Affymetrix mouse MoGene 1.0 platform (Affymetrix). From a total of 28,188 annotated probesets, mean differences in gene expression were statistically identified by applying a 2-way ANOVA block design using the Linear Model for Microarray Data (Limma) Package47,48 through implementation of MicroArray Я US Gui interface to R49 and the Bioconductor Project microarray analysis programs. The microarray data has been assigned a GEO omnibus accession GSE95579.

Cochlear function

Auditory brainstem responses (ABRs) and distortion product otoacoustic emissions (DPOAEs) were recorded in WT, Tg1MDW hemizygous and Tg1MDW/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 previously50,51. 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 Kjær 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.

DPOAE Procedures

DPOAEs were used to assess OHC function non-invasively as detailed previously51,52. 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, Tg1MDW hemizygous and Tg1MDW/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.