An ENU-induced mutation of miR-96 associated with progressive hearing loss in mice

Progressive hearing loss is common in the human population. About one in 850 children are born with a significant, permanent hearing impairment, but by the age of ten this number has doubled¹. Age-related hearing loss in later life has a heritability approaching 50%², and some single genes have been identified underlying progressive hearing loss in rare families (Hereditary Hearing Loss Homepage; http://webho1.ua.ac.be/hhh/). However, for the vast majority of cases of progressive hearing loss there is no molecular diagnosis. To provide candidate genes and models for hearing loss, we established a screen for new ENU-induced deaf mouse mutants³. One such mutant recovered was diminuendo (Dmdo), inherited in a semi-dominant manner. Heterozygotes (Dmdo/+) show a progressive loss of the Preyer reflex (ear flick response to sound) between 4 and 6 weeks. Homozygotes (Dmdo/Dmdo) do not demonstrate a Preyer reflex at any age, and show head bobbing and a staggering, circling gait.

The gross structure of the middle and inner ears appeared normal in mutants so we examined the organ of Corti using scanning electron microscopy. At 4-5 days after birth, the number and arrangement of hair cells appeared normal in mutants, and heterozygotes looked similar to wildtype littermates (Fig 1a, S1a,b). However, irregular bundles and persistent clusters of ectopic stereocilia were observed in homozygotes (Fig. 1b-d) and by 7 days, homozygote hair cells displayed marked degeneration (data not shown). At four and six weeks old, very few recognisable hair cells remained in homozygotes (Fig. S1f,k). In heterozygotes many outer hair cells had degenerated in the middle and basal turns but most inner hair cells remained intact (Fig S1l-o). Remaining outer hair cell stereocilia bundles formed a loose U-shape rather than the precise V-shape of controls, and inner hair cells often showed smaller, more widely-separated bundles (Fig. 1e,f,g,h; S1e,h,j,p,q). Compound action potentials, reflecting cochlear nerve activity in response to sound, were undetectable in homozygotes at 4 weeks, and in heterozygotes thresholds were raised by around 60dB (Fig. S1r), despite the persistence of many surviving hair cells. Endocochlear potentials were within the normal range (Fig. S1s). Fused stereocilia and hair cell degeneration were evident in the vestibular systems of four week-old heterozygotes.

We mapped the mutant phenotype to proximal chromosome 6, between D6Mit159 and D6Mit268, a 4.96Mb interval (Fig S2a,b). We sequenced 87% of the ~900 exons within the interval (http://www.ensembl.org), and located two mutations. The first mutation was a silent C>T substitution in exon 5 of 2310005E10Rik, a member of the aldo-keto reductase family and a putative orthologue of the human gene AKR1B10. The mutated base is the third in the codon, and the amino acid, asparagine, is not affected. The variant in the Dmdo genome is identical to the equivalent wildtype human reference sequence. Normalised cDNA from the organs of Corti of three P4 Dmdo/Dmdo and +/+ sibling pairs gave bands of identical size and intensity when subjected to PCR with primers in exons 4 and 6. We concluded that this variant was unlikely to be involved in causing the hearing impairment in the diminuendo mutant (Fig S2c,d).

The second mutation was an A>T substitution in Mirn96 (mmu-miR-96; miR-96), a microRNA (miRNA). MicroRNAs are small non-coding RNAs that bind to specific sites in the 3′UTR of target mRNAs, inducing transcript destabilisation and translational inhibition. The mutation lies within the miRNA seed region (nucleotides 2-7), which confers binding specificity⁴, making this variant a strong candidate for the causative mutation. The sequence of mature miR-96 is perfectly conserved between human, mouse, rat and fish (Fig. 2a).

Mirn96 is one of a cluster of three miRNAs; the other two are Mirn183 and Mirn182. These miRNAs are expressed in all sensory hair cells of the inner ear and in neurons of the cochlear and vestibular ganglia in newborn mice, and are thought to be processed from a single primary transcript⁵. We found that all three mature miRNAs were expressed in hair cells of both wildtypes and homozygous mutants (Fig. 2b-g), suggesting the mutation does not have a major effect on processing and export of the miRNAs to the cytoplasm despite the possible interference of the mutation with stem loop formation required for cleavage of the transcript⁶.

We chose two complementary approaches to study the effects of the mutation. Firstly, we used the miRanda target predictor v3.0⁷with additional filtering to produce a list of 132 genes which we annotated (Supplementary Table 1) and selected thirteen for further study (Supplementary Table 2). We used a luciferase assay with siRNAs mimicking the wildtype and mutant miR-96. The siRNAs were co-transfected into NIH 3T3 cells with a construct containing the 3′UTR of each candidate gene, either with the putative binding sites intact, or with the sites replaced by an EcoR1 site to disrupt binding. Five genes were validated as targets of miR-96: Aqp5, Celsr2, Myrip, Odf2 and Ryk (Fig. S3a,b). Quantitative RTPCR showed Aqp5 and Celsr2 were significantly upregulated in mutant cochlear tissues compared with wildtype. However, the difference in expression levels was small (Fig. S3c). We used antibodies against the validated targets and found all five were expressed in or near wildtype hair cells at P3 and P5, but there was no visible difference in diminuendo (Fig S3d-m and data not shown). However, miRNAs may have multiple small effects on the expression levels of a number of genes⁸ and immunohistochemical tests may not show such small effects. Therefore, we adopted a genome-wide approach to investigate the mechanism of action of the mutation.

We compared gene expression of both direct and indirect targets by microarray analysis of the organ of Corti of P4 mutants and wildtypes. We retrieved 96 significantly affected transcripts (P-value<0.05); 50 genes were up-regulated and 36 down-regulated (Supplementary Table 3; the remaining 10 probes were either duplicates (6) or mapped to intergenic regions (4)). Thirteen of these so far have been confirmed by qRTPCR (Fig. S4a). Of the downregulated genes, five in particular were of interest; Slc26a5 (prestin), Ocm (oncomodulin), Pitpnm1, Gfi1 and Ptprq. None of these genes have a predicted wildtype or mutant miR-96 target site, so the downregulation is presumably a downstream effect. Prestin is a voltage-sensitive motor molecule that mediates outer hair cell length changes responsible for amplification of sound within the cochlea⁹. Prestin knockout mice have short hair cells and hair cell degeneration¹⁰. Oncomodulin is expressed in outer hair cells and may act as a cytosolic calcium ion buffer¹¹. Ptprq is required for maturation of the hair bundle, and is thought to be a component of interstereocilial shaft connectors¹². Gfi1 is expressed in hair cells, and knockout mice display hair cell degeneration¹³. The difference in expression of these genes was confirmed by qRTPCR in both heterozygotes and homozygotes (Fig. 3a) and by immunohistochemistry (Fig. 3b-k). We found no evidence of genomic changes that might account for the extreme downregulation of Ocm and Slc26a5: exon 1 of each gene amplified correctly in mutants, surrounding genes were expressed normally in the microarray, and the diminuendo phenotype showed Mendelian inheritance and mapped to the Mirn96 locus. Epigenetic downregulation of any one of these five genes could explain the hearing impairment, as three are known to lead to deafness when knocked out and the remaining two are highly expressed in sensory hair cells.

We asked whether the striking downregulation of oncomodulin and prestin was a generic feature of degenerating hair cells by looking at immunostaining intensity in nine other mouse mutants which exhibit early hair cell degeneration: headbanger and shaker1^4626SB (Myo7a)^14,15, Snell's waltzer (Myo6) ¹⁶, headturner (Jag1)¹⁷, beethoven and deafness (Tmc1) ^18,19, oblivion (Atp2b2) ²⁰, catweasel (Six1)²¹ and whirler (Whrn)²². Oncomodulin and prestin showed hair cell labelling in mutants as strong as in the littermate controls (Fig. S5a-p and data not shown), suggesting that the reduction of expression in diminuendo was a specific feature. Furthermore, other markers of hair cells and supporting cells of the organ of Corti showed normal immunostaining intensity in diminuendo mutants at P0, P3 and P5, including Myo7a, Cdkn1b (p27^Kip1), Sox2 and Jag1 (Fig S5q-y and data not shown).

We next searched for wider miRNA effects on the mRNA profile of diminuendo using Sylamer²³. Analysis of all miRNA heptamers shows that the heptamer complementary to the seed region of miR-96 (GUGCCAA) is greatly enriched in the 3′UTRs of hundreds of genes upregulated in diminuendo homozygotes (Fig. 4). This indicates that miR-96 normally modulates expression of a broad range of target genes, and that it affects mRNA levels rather than affecting translation alone. Among the most downregulated genes, the heptamer complementary to the mutant miR-96 is enriched (GAGCCAA, Fig. 4), indicating that mutant miR-96 influences expression of newly-acquired target genes. We analysed conservation of these signals. Wildtype seed matches are enriched in 3′UTRs of human and rat orthologues of the most upregulated mouse genes (Fig 4), suggesting that these sites are conserved and may be functional. However, enrichment of mutant miR-96 binding sites in human and rat orthologues of downregulated genes is barely above significance threshold (dotted line). We also employed a simple log ratio analysis of seed match biases in gene sets at different cutoffs (Fig S4b-j). These results agreed with the Sylamer analysis.

To elucidate the link between the mutated microRNA and the abnormally-expressed genes revealed by the microarray analysis, we examined the 500bp upstream of the top 356 upregulated genes and the top 425 downregulated genes to find any transcription factor motifs that were enriched in the affected genes (Fig S4k-n). Several interesting motifs were found, including a Gfi1-like motif found among the upregulated genes; a binding site for Mitf, known to be regulated by miR-96 ²⁴; and targets associated with control of Notch signalling, such as Pou2f1²⁵, Rbpj²⁶, and bHLH transcription factors²⁷. Any one of the transcription factor binding motifs discovered could be involved in linking the mutation in miR-96 to the misregulation of genes directly required for hair cell development and survival. The large number of genes whose expression is affected by miR-96 and the complexity of their interactions suggests that there may not be a simple mechanism that explains the effects of the mutation, but rather a combination of many small effects that act in concert to cause hair cell dysfunction.

The diminuendo mutant shows progressive hearing impairment in heterozygotes and profound deafness in homozygotes associated with hair cell defects. Although hearing impairment is often thought to be caused by hair cell loss, this and previous studies suggest that the degeneration is instead a correlate or consequence of a prior dysfunction of the hair cells. The mutation in the seed region of miR-96 is highly likely to cause the hearing impairment in diminuendo mutants because it co-segregates with the phenotype, it occurs in the seed region of the miRNA known to be critical for target recognition, miR-96 is expressed specifically in the cells most affected by the mutation, Sylamer analysis indicated that the mutation has a direct effect on expression of many genes as well as indirect effects, and we found no other plausible mutation despite resequencing the vast majority of the coding sequence within the non-recombinant region of chromosome 6. In addition, the finding of two different single base changes in the seed region of miR-96 in humans with progressive hearing loss in the accompanying report²⁸ provides critical support for our proposal that the single base change in miR-96 is the causative mutation behind the diminuendo phenotype, and furthermore suggests that the phenotype results from a lack of repression of normal targets even though we show a gain of repression of novel targets. Although the link between the direct targets of the miRNA and the phenotype are not clear, we have shown that several genes known to be important for hair cell function are specifically downregulated in the diminuendo mutant and any one could account for the hair cell dysfunction. This is the first ENU-induced mutation found in a miRNA and the first miRNA found to be associated with deafness. Understanding the mechanism by which miR-96 leads to progressive hearing loss will give us clues to help develop therapies to ameliorate the effects of progressive deafness, whatever the trigger.