An L1 retrotransposon insertion–induced deafness mouse model for studying the development and function of the cochlear stria vascularis

Significance

To gain better hearing sensitivity, the energetic and metabolic supplies are crucial for cochlea to develop a blood-labyrinth barrier containing abundant capillaries and cellular compartments. For example, the stria vascularis, the power plant of the cochlea, generates a +80 mV endocochlear potential and a 150 mM potassium concentration in the scala media. Previous studies of stria vascularis development and function have been largely focused on characterizing transcription factors, channels, and transporters. In our study, we identify a retrotransposon insertion mutation responsible for complete stria vascularis loss of function. Our study further highlights the diversity of genetic mutations capable of inducing inherited deafness.

Abstract

Dysregulation of ion and potential homeostasis in the scala media is the most prevalent cause of hearing loss in mammals. However, it is not well understood how the development and function of the stria vascularis regulates this fluid homeostasis in the scala media. From a mouse genetic screen, we characterize a mouse line, named 299, that displays profound hearing impairment. Histology suggests that 299 mutant mice carry a severe, congenital structural defect of the stria vascularis. The in vivo recording of 299 mice using double-barreled electrodes shows that endocochlear potential is abolished and potassium concentration is reduced to ∼20 mM in the scala media, a stark contrast to the +80 mV endocochlear potential and the 150 mM potassium concentration present in healthy control mice. Genomic analysis revealed a roughly 7-kb-long, interspersed nuclear element (LINE-1 or L1) retrotransposon insertion on chromosome 11. Strikingly, the deletion of this L1 retrotransposon insertion from chromosome 11 restored the hearing of 299 mutant mice. In summary, we characterize a mouse model that enables the study of stria vascularis development and fluid homeostasis in the scala media.

Hearing sensitivity is largely determined by the exquisite structure and function of the cochlea, which converts sound information into encoded impulses in auditory nerves (1). The conversion is performed by cochlear hair cells via the mechanoelectrical transduction (MET) channels located in their hair bundles (2). Intriguingly, hair bundles are immersed in a sealed space within the cochlea called the scala media, which exclusively contains endolymph fluid (3, 4). To power MET activity, the scala media provides a high level of endocochlear potential (EP) (+80 mV) and a high-potassium concentration (∼150 mM), both of which are produced and maintained by the stria vascularis (StV) positioned at the inner side of the lateral wall (LW), facing the scala media (3, 4).

In order to secrete endolymph into the scala media and maintain the fluid homeostasis, including the EP and K⁺ concentration, the StV develops a sandwich-like structure, consisting of three predominant cell types. They are marginal cells (MCs), intermediate cells (ICs; melanocyte like), and basal cells (BCs), and a capillary network is distributed between the MCs and BCs (5, 6). Within the StV, each cell type has specific functions that are dictated by the expression of specific genes, including the following: tight junction, gap junction, ion channel, and transporter genes (6). Given the important functions of the StV, mutations in StV-expressed genes cause abnormal fluid homeostasis in scala media and syndromic or nonsyndromic sensorineural deafness (6).

Among StV-associated genes, transcription factors critically regulate the development and maintenance of the StV. Esrrb (also known as Nr3b2) mainly regulates gene expression profiles in MCs, which further modulates inner ear development. Following Esrrb deficiency, inner ear development is impaired with a volume-reduced scala media and a narrow semicircular canal (7). Furthermore, Mitf, Sox10, and Pax3 are important gene expression regulators in ICs, along with their Waadenburgh syndrome and melanocyte pathology functions. Mutations in these genes induce profound damage or the loss of ICs, giving rise to melanocyte-related disorders, including hearing loss (8–11).

Here, we describe a nonsyndromic congenital deafness mouse line, 299, that displays a structural defect of the StV. We identify the 299 causative mutation as an insertion of a transposable element, long interspersed element-1 (LINE-1 or L1), within an intergenic region on chromosome 11 (chr11). Interestingly, this genetic variation has not been previously associated with deafness (12). As such, the 299 mouse line is a suitable animal model for studying the StV development and function and fluid homeostasis mechanisms in the inner ear, as well as the pathogenic mechanism of retrotransposon-mediated genomic regulation.

Results

Characterization of the Deaf Mouse Line 299.

The 299 mouse line was originally generated from a forward genetic screen of deaf mice induced by N-ethyl-N-nitrosourea (ENU) mutagenesis (13). We first characterized 299 mice hearing capability, by examining their auditory brainstem response (ABR) thresholds at 3 to 4 wk of age. The ABR assay reflects general hearing sensitivity by measuring auditory nerve discharges at multiple layers from the cochlea to the brainstem (14). Click stimuli that represent white noises did not evoke ABR waveforms in 299 mice, even at 90 dB sound pressure level (SPL) (the highest intensity applied), while the control wild-type (WT) mice showed ABR signals at 30 dB SPL and exhibited the typical five major waves (Fig.1 A and B). Similarly, 299 mice also showed the complete loss of pure-tone ABR signals following tone stimuli at all tested frequencies (4 to 32 kHz) (Fig. 1C). Compound action potential (CAP) recordings were captured by placing an electrode in the facial nerve canal, near the auditory nerves (15). No click-induced CAP signals were recorded in 299 mice, while control WT mice displayed expected CAP signals S1A and S1B. As the vestibular epithelium and the cochlea typically express many of the same genes, we further examined the balance function and motor coordination capabilities of 299 mice using a rotarod test (16). We find that 299 mice at 2 mo of age retain their balance function, relative to WT control mice (SI Appendix, Fig. S1 C–E). No other obvious health problems or phenotypes were observed in 299 mice. Our results suggest that 299 mice harbor a profound, likely nonsyndromic, inherited deafness.

Fig. 1.

Analysis of auditory function of 299 mouse line. (A) Representative examples of click ABR signals from a WT control mouse and a 299 mouse. The click sound was delivered at graded intensities (at 10-dB step). The 299 mouse shows ABR thresholds >90 dB SPL. (B and C) Quantification of click (B) and pure-tone (C) ABR thresholds in WT control and 299 mice. Thresholds of 299 mice are elevated to over 90 dB SPL at all tested frequencies (arrow). Unpaired t test, ****P < 0.0001 and error bars, SD. (D) Representative examples of DPOAE response spectra in a WT mouse and a 299 mouse at 12 kHz 70-dB SPL simulation. The 2F1 to 2F2 distortion product (DP) presents in the WT mouse but not in the 299 mouse. (E) Comparison of DPOAE thresholds of WT and 299 mice. DPOAE thresholds of 299 mice are elevated to over 80 dB SPL at all analyzed frequencies. Unpaired t test, 4 kHz, ns (no significance), P > 0.05; 8 kHz, and *P = 0.038; 12, 16, and 20 kHz, ****P < 0.0001; 32 kHz, *P = 0.014; and error bars, SD. In A–E, all recordings were obtained from mice at the age of 3 to 4 wk. In B, C, and E, black arrows indicate that the mean thresholds are higher than shown. n numbers and frequencies of stimuli are shown in panels.

We next used the distortion product otoacoustic emission (DPOAE) assay to examine the contribution of cochlear hair cells to the deafness of 299 mice. DPOAE is a noninvasive audiometry approach, in which DPOAE thresholds at a given frequency reflect the active amplification function of the corresponding outer hair cells (OHCs) (17). In 3 to 4 wk old 299 mice, DPOAE thresholds at all tested frequencies were elevated to 80 dB SPL (Fig. 1 D and E). These lack of DPOAE signals suggest a loss of OHC motility function in 299 mice, which may contribute to their profound deafness. From these results, we next examined the maintenance of cochlear hair cells in P21 299 mice by immunostaining. We found that the number of OHCs in 299 mice was drastically reduced, following a gradient from apex to base (Fig. 2 A and B), while inner hair cells (IHCs) were not affected. Along the cochlea, OHC survival rates were 13.6% at base, 45.1% at midst, and 96.6% at apex (SI Appendix, Fig. S1F). These results suggest that a major cause of deafness in 299 mice is likely due to the loss of OHCs.

Fig. 2.

Analysis of hair cells in 299 mice. (A and B) Immunostaining of the whole-mount cochleae from WT (A) and 299 (B) mice at P21. The cochleae were stained with Myo7a antibody (white) and phalloidin (red). OHCs at middle and basal coils in 299 mice degenerate gradually, while IHCs are intact. (Scale bar, 20 μm.) (C) SEM photographs showing middle coils of cochleae from WT and 299 mice at P21. (Bottom) Enlarged photographs (framed with yellow dashed line) show the loss of hair bundles in the 299 mice. (Top: scale bar, 20 μm; Bottom: scale bar, 10 μm.) (D) SEM photographs of cochleae from WT and 299 mice at P7. At apical, middle, and basal coils, the hair bundle morphology of all OHCs is normal in 299 mice. (Scale bar, 10 μm.) (E) Representative traces of MET currents from OHCs of WT and 299 mice at P7. The OHC was recorded at a whole-cell patch clamp configuration, and the hair bundle was stimulated with a series of deflection from −0.4 to 0.9 μm at 0.1-μm step (Upper). (F) Quantification of the MET current and open probability from recordings similar to E. Unpaired t test, ns, P > 0.05; n numbers and P values are shown in panels; and error bars, SD.

We next wondered whether OHC loss in 299 mice is a consequence of MET defects, as MET is essential for hair–cell functions, and MET malfunction induces hair–cell degeneration in the long term (18). Scanning electron micrograph (SEM) experiments in P21 299 mice showed that surviving OHCs maintain their hair bundles with a relatively normal morphology (Fig. 2C). P7 299 mice also showed no loss of hair cells and normal hair bundles (Fig. 2D). To further characterize MET functions in OHCs, we recorded hair bundle deflection–activated currents in P7 299 mice (Fig. 2 E and F). Surprisingly, MET current was slightly increased instead of being abolished or reduced in homozygous 299 mice, as compared to heterozygous control mice (Fig. 2 E and F). However, we observed no change in the open probability of MET channels in homozygous 299 mice (Fig. 2F, Right). These results suggest that 299 mice preserve the MET function, at least in the first neonatal week.

Stria Vascularis Disruptions in 299 Mice.

Previous reports suggest that StV deficiency leads to hair cell death (19). We speculated that loss of OHCs in 299 mice at P21 might be resulted from an environmental change in the organ of Corti, for example, fluid homeostasis in the scala media (19). To test this hypothesis, we carried out hematoxylin and eosin (HE) staining on paraffin sections of the inner ears from P21 299 mice. In these 299 mice, we found the complete loss of the StV (Fig. 3 A and B and SI Appendix, Table S1) and the Reissner’s membrane (RM) attached directly to the spiral prominence (SP), with the scala media space significantly reduced (Fig. 3A, Bottom and Fig. 3B). In contrast, no obvious defects were observed in the spiral ganglia (SG) (Fig. 3A, arrows), as quantified by the number of SG neurons (SI Appendix, Fig. S2 A and B). These results suggest that the auditory deficiency in 299 mice is due to the StV disruption, as opposed to an auditory nerve deficiency.

Fig. 3.

Abnormal morphology of StV in 299 mice. (A) HE staining of cochleae from WT and 299 mice at P21. (Left) StV was lost completely along the cochlear coil in 299 mice, without other obvious defects found in the cochlear structure. (Middle and Right) Enlarged photographs showing the StV structure as numbered in the left panels. (Left: scale bar, 200 μm; Middle and Right: scale bar, 20 μm.) (B, Top) Diagram showing the way to quantify StV loss from HE staining pictures similar to A. Based on the position that RM attaches to SL, the length of SL was split as Lx and Ly. The ratio, Lx/(Lx + Ly), was calculated. In addition, the volume change of scala media (SM) was calculated by comparing the area of SM (A_SM) with the area of scala vestibuli (SV) (A_SV). (Bottom) The quantification data. Only the third organ of Corti was counted from each cochlea. Unpaired t test, ****P < 0.0001; n numbers are shown in panels; and error bars, SD. (C) SEM of the inner side of the LW. The cochleae were collected from WT and 299 mice at P21. (Upper: scale bar, 100 μm; Middle: scale bar, 20 μm; Lower: scale bar, 10 μm.) (D) Immunostaining of the StV and SL capillaries in the LW from a P7 WT mouse (Upper) and a 299 mouse (Lower). Capillary was labeled by CD31 antibody (red). (Scale bar, 200 μm.) The abbreviation in the panels is the following: BM, basilar membrane.

We further examined the vestibular end organs of 299 mice. No obvious defects were found, as noted by the dark cells in ampullae (SI Appendix, Fig. S2C) and hair cells with otoconia in saccules and utricles (SI Appendix, Fig. 2D). These findings are consistent with our physiological data that 299 mice do not have impaired balance functions (SI Appendix, Fig. S1 C–E).

To further characterize the StV defects in 299 mice, we performed SEM to inspect MC morphology. In P21 299 mice, MCs were lost completely, while the RM moved to a lower position and attached directly to the SP (Fig. 3C). We observed this same pattern at P7 (SI Appendix, Fig. S3A). In the first neonatal week, LW vessels form a single-layer capillary network at birth and are subsequently divided into two layers: StV capillary and spiral ligament (SL) capillary, respectively (20). We used a CD31 antibody (21) to label capillary endothelial cells. In line with our HE staining and SEM observation, we found that the capillary StV layer in 299 mice was completely lost, and the capillary SL layer became irregular at P7 (Fig. 3D). Furthermore, disordered vasculature was observed as early as P1 (SI Appendix, Fig. S3B), suggesting that 299 mice have an early developmental disruption of the LW vasculature. These results provide further evidence for 299 mice having a severe StV defect.

As P21 is an age when mice have a mature cochlear structure, we next investigated the progression of StV morphology in 299 mice at different developmental time points. We performed HE staining on of the paraffin sections of 299 cochleae at E18, P1, P4, P7, and P14 (SI Appendix, Fig. S4 and Table S1). Consistent with WT control mice, E18 299 mice have a single-cell layer in the StV from the basal to apical turn, as in WT mice (SI Appendix, Fig. S4A). In WT P1 mice, the StV then develops into multiple cell layers in the basal and middle turn. However, in P1 299 mice, the StV remains as a single-cell layer from the basal to apical turn (SI Appendix, Fig. S4B). In P4 WT control mice, three layers of cells then emerge clearly, and the RM connects to the upper edge of the StV. However, in P4 299 mice, we observed no StV and only SP, with the RM connecting directly to the SP (SI Appendix, Fig. S4C). At P7 and P14, 299 mice also lack an StV (SI Appendix, Fig. S4 D and E). These results reveal that 299 mice completely fail to develop the StV at all. Thus, the 299 mice are a suitable animal model for studying StV development, including the aspects of layer specialization and the formation of the capillary network.

EP and K⁺ Concentrations Are Disrupted in 299 Mice.

As the StV secrets K⁺ into the endolymph and generates EP for the scala media (22), we recorded EP in the scala media, as previously described (23). While inserting a sharp recording electrode through the scala tympani, basilar membrane, scala media, and scala vestibuli sequentially (Fig. 4A), we recorded potential changes as the electrode advanced. In both WT and 299 mice, when the electrode penetrated the organ of Corti, we observed an ∼−70 mV potential (Fig. 4B), known as organ of Corti potential (24). Subsequently, in WT mice, EP increased to about +90 mV after the electrode entered the scala media and returned to 0 mV until the electrode entered the scala vestibuli (Fig. 4B, Upper). In contrast, we recorded no EP in the scala media of 299 mice (Fig. 4B, Bottom). These results show that EP is abolished in 299 mice (Fig. 4C), indicating that the endolymph ion homeostasis in the scala media might change. As potassium is the dominant ion within the scala media, we performed a double-barreled, electrode-recording technique to measure K⁺ concentrations in vivo, following a penetration path similar to that in Fig. 4A. Compared to the 150 mM K⁺ concentrations in WT control mice, the K⁺ concentrations in the scala media of 299 mice reduced dramatically to only ∼20 mM (Fig. 4 D and E), in line with the EP defects (Fig. 4C). These results indicate that the StV deficiency in 299 mice induces the loss of EP and K⁺ and further suggests that OHC death is likely a consequence of StV deficiency (19).

Fig. 4.

Disrupted EP and [K⁺] in the scala media (SM) of 299 mice. (A) Procedure of EP measurement. A sharp electrode was entered from the scala tympani (ST), passed through the SM, and finally reached the scala vestibuli to measure the potential during the process. (B) Example traces of the measured potential in WT and 299 mice at 3 to 4 wk. Note that EP is ∼+90 mV in the SM of the WT mouse while it’s 0 mV in the 299 mouse. (C) Quantification of EP from recordings similar to B. n numbers are shown in panels. Unpaired t test, ****P < 0.0001 and error bars, SD. (D) Representative examples showing the simultaneous EP and [K⁺] recording traces from a WT mouse and a 299 mouse. Similar to the punctuation steps in A, a double-barreled K⁺‐selective electrode was advanced through cochlear tube as indicated (filled arrows). EP and [K⁺] in the SM of 299 mice decrease dramatically. Note that in the 299 case an abruptly high-potassium concentration (arrow), when penetrating basilar membrane (BM), may reflect an encounter of the electrode with cell cytosol. (E) Quantification of [K⁺] in the SM from recordings similar to D. n numbers are shown in the panel. Unpaired t test, ****P < 0.0001 and error bars, SD.

We suspect that the observed slight increase in MET current amplitude may be a compensatory mechanism due to the lack of EP. Previous reports suggest that OHC MET current represents the tonotopic (both temporal and spatial) characteristics, which increase from apical OHCs to basal OHCs (25). We explored whether OHC tonotopic features are affected by StV loss and are established by EP and K⁺ (26). To do this, we tested the maximal MET current in OHCs at the different positions along the cochlea (distance to the apex, 20, 40, 60, and 80%) at P7 (SI Appendix, Fig. S5 A and B). We found that 299 mice maintain a tonotopic pattern of maximal MET currents, and these currents generally increase at all the tested spatial positions than those in control mice (SI Appendix, Fig. S5 A and B). We further examined OHC MET currents at comparable positions (40% to the apex) of the basilar membrane in P1, P4, and P7 WT and 299 mice. We found that OHC MET currents in 299 mice increase in the first neonatal week (SI Appendix, Fig. S5 C and D). Consistent with the increased MET current amplitude, the number of stereocilia in OHCs instead of IHCs increased from apex to base along the cochlear coil in both WT and 299 P7 mice (SI Appendix, Fig. S5 E–H). We also found that P7 299 OHCs have more stereocilia than WT control OHCs, especially at the apex (SI Appendix, Fig. S5 E and F). These results demonstrate that the tonotopic dependency of MET current is preserved in 299 mice, and reduced EP and K⁺ in the scala media likely increase the number of stereocilia in OHCs, which may give rise to the increased MET currents observed in 299 mice during the first neonatal week.

No Pathogenic Mutations Were Found by Next-Generation Sequencing.

To explore the inheritance pattern of 299, we outcrossed a deaf 299 mouse with an ICR WT mouse. G1 offspring were self-crossed to obtain the G2 generation. Click ABR thresholds of all G2 offspring were measured, indicating that 299 deafness follows a recessive, Mendelian-inheriting pattern (SI Appendix, Table S2). Prior positional cloning experiments had identified a chr11 interval (chr11: 16,900,000 to 18,310,000) that likely contained the mutation(s) responsible for deafness in 299 mice (13). As the 299 line was initially found from an ENU screening project and ENU preferentially induces A-to-T transversions, we first hypothesized that the deafness-causing mutation in 299 mice may be a small-scale mutation, such as a single nuclear polymorphism (SNP) or insertion/deletion (InDel) (27–29). However, we identified no SNP/InDel mutations using Sanger sequencing of all exons in the nine coding genes of the chr11 candidate interval (SI Appendix, Tables S3 and S4). To more comprehensively identify the causative mutation in 299 mice, we performed short-read, whole-genome next-generation sequencing (NGS) on a 299 homozygous 299 mouse and a heterozygous littermate control mouse. In our NGS analysis of the chr11 candidate interval (SI Appendix, Fig. S6A), we identified only one T > C SNP within an intron of the Ppp3r1 gene (SI Appendix, Fig. S6 B and C). We then amplified and Sanger sequenced this T > C mutation in all 299 G1 and G2 offspring, confirming that G1 299 mice are heterozygous for this mutation, with each mouse carrying both a T and C allele at this position (SI Appendix, Fig. S6D). We also confirmed that the phenotype and T > C mutation were matched in all 299 mice (SI Appendix, Fig. S6E and Tables S2 and S5). To test whether this T > C mutation is linked to 299 mouse deafness, we generated a knock-in (KI) (Ppp3r1-KI) mouse line harboring this mutation and obtained F2 Ppp3r1-KI homozygous mice by self-crossing heterozygous F1 mice (SI Appendix, Fig. S6F and Table S6). However, click ABR thresholds (SI Appendix, Fig. S6 G and H) and the StV morphology (SI Appendix, Fig. S6I) of homozygous Ppp3r1-KI mice were normal and matched that of heterozygous controls. These results suggest that this Ppp3r1 T > C mutation is not causative for the deafness observed in 299 mice.

We also analyzed the sequences of StV-related genes, such as Esrrb, Nkcc1, and Mitf in the NGS data, as these genes are tightly related to the development or function of StV (7, 8, 30–32). However, we identified no mutations in any of these genes. From these short-read NGS experiments, we were unable to find the deafness-causing variant in 299 mice, which had been considered as the consequence of ENU-based mutagenesis.

PacBio Sequencing Identifies an L1 Retrotransposon Insertion.

As typical NGS produces a short-read (∼150 to 300 bp) library (33, 34) and can have low- or nonexistent read coverage at highly repetitive and complex genome sequences, we further expanded our search for the deafness-causing variant in 299 mice to include large-scale and structural mutations. We performed long-read PacBio sequencing in 299 mice, which is capable of detecting large structural variants (35). Similar to a previous study (36), most structural variants in 299 mice were insertions in intergenic regions (Fig. 5 A and B). PacBio sequencing results revealed four PacBio structural variants (PBSVs) within the chr11 interval (Fig. 5C). To validate these four candidates, we viewed aligned reads in the Integrative Genomics Viewer (IGV) (SI Appendix, Fig. S7 A–D) and PCR amplified them from both WT and 299 mouse genomic DNA (SI Appendix, Fig. S7E). Specifically, subreads aligning to the PBSV2304 insertion site (chr11: 17,614,267) were consistent to the insertion pattern observed in the PacBio data (Fig. 5D). We found that an ∼7 kb PCR product was amplified in 299 genomic DNA but not in WT mouse genomic DNA (Fig. 5E), consistent with the 7,051-bp insertion (PBSV2304) observed in the PacBio data (Fig. 5 C and D). We sequenced the ∼7-kb fragment after inserting it into the pcDNA3.1 plasmid vector. The sequence aligned to a nearly full-length L1 retrotransposon in chromosome 14 (chr14: 10,489,690 to 10,496,741). Upon viewing the PacBio data in IGV, we noticed that the PBSV2304 insertion also displayed a 15-bp duplication at the insertion site (Fig. 5D), a target site duplication event that occurs during L1 retrotransposition (37). From this, we hypothesized that the pathogenic mutation of 299 mice may be an L1 insertion (L1-INS) induced by a retrotransposition event.

Fig. 5.

Identification of an L1 retrotransposon insertion in chr11 of 299 mice. (A) Distribution of structural variant locations acquired by PacBio sequencing in a 299 mouse genome. Most structural variants are located in the intergenic regions. (B) Distribution of structural variant types. Over 90% of detected structural variants are insertions. (C) Four candidates of PBSVs in the previously determined pathogenic interval at chr11 (chr11:16,900,000 to 18,310,000). The PBSV2304 (a 7,051-bp insertion) was considered as the top candidate (red). DUP, duplication; INS, insertion. (D) Diagram visualizing the genes and PBSV calls in the chr11 interval. The nine coding genes are shown at the corresponding location. The four candidates of PBSVs are shown by vertical lines. IGV view of PacBio subreads of the top candidate (PBSV2304) is shown at the bottom. A near, full-length (7,051 bp) L1 element (from chromosome 14) was found inserted into the chr11 interval. Forward and reverse complement strands are shown in pink and blue, respectively. A 15 bp target site duplication (TSD) was found in PBSV2304. (E) Verification of PBSV2304 from WT and 299 genomes by PCR. An ∼7-kb size band was amplified from genomes of two individual 299 mice compared to that from WT mice. M, markers.

The L1-INS Causes Inherited Deafness of 299 Mice.

If our identified L1-INS causes 299 mice deafness, we hypothesized that deleting one L1-INS allele in homozygous 299 mice will rescue the mutant phenotype. Therefore, we designed two single guide RNAs (sgRNAs) targeting the flanking sequences of the 299 L1-INS and used a Cas9 endonuclease to specifically remove the L1-INS sequence from the 299 genome. A 120-bp single-strand oligonucleotides (ssODNs) was used to bridge the double-strained breaks following deletion of the L1-INS (Fig. 6A and SI Appendix, Table S7) (38). We obtained 19 F1 offspring after injection and identified seven mice (#2, 3, 6, 7, 10, 11, and 12) with possible deletion events, according to our genotyping (Fig. 6B and SI Appendix, Fig. S8A). We also obtained one mouse (#10), for which both L1-INS alleles were deleted (SI Appendix, Fig. S8A, box in yellow dashed line). Click ABR recordings showed that hearing was restored in five L1-INS–deleted mice (#2, 6, 10, 11, and 12) (Fig. 6C). Of the seven genotyped mice, #3 and #7 displayed the partial rescue of the hearing phenotype (SI Appendix, Fig. S8 B and C) and were confirmed as mosaic L1-INS deletions, following the ABR examination of their offspring crossed with 299 mutant mice (SI Appendix, Table S8). Phenotype-rescued 299 mice with L1-INS deletion (named 299KO) were further outcrossed with 299 mutant mice, and the offspring were tested by click ABR (Fig. 6E). We found that ABR threshold phenotypes of all offspring matched L1-INS genotypes (SI Appendix, Table S9). Strikingly, the StV morphology in 299KO mice was also rescued (Fig. 6D). These results indicate that the identified L1-INS at chr11: 17,614,267 is the pathogenic and deafness-causing mutation in 299 mice.

Fig. 6.

Deletion of the L1-INS from chr11 restores 299 mouse hearing. (A) Diagram showing the generation of 299KO mice by deleting the L1 element from chr11 in 299 mice. Double sgRNAs, Cas9 messenger RNA, and ssODN were mixed and injected into pronucleus of the 299 homozygous zygotes. Ref Seq, reference sequence and TSD, target site duplication. (B) Diagram of corresponding sequences of 299 mutant allele and 299KO allele. The sequences of sgRNAs (underlined) and protospacer adjacent motif (PAM) (bold) are shown. In 299KO allele, a 7,137-bp fragment is deleted compared to 299 mutant allele. After deleting the L1 element, only an 8 bp sequence remains. (C) ABR traces of WT control and 299KO F1 mice. Both #2 and #10 299KO mice had normal hearing after L1 were deleted in single (299KO-Heter #2) or both (299KO-Heter #10) mutant alleles. Without the deletion of the L1, the ABR threshold of the #1 299 mutant mouse was still >90 dB SPL. (D) HE staining of cochleae in 299KO heterozygous mice (299KO-Heter). The StV of 299KO-Heter mice is as normal as WT control. BM, basilar membrane. (Scale bar, 50 μm.) (E) ABR tests of F2 mice with L1 deletion. The F2 299KO-Heter mice were generated by crossing the 299KO F1 mice with the 299 homozygous mice. n numbers are shown in panels. Unpaired t test, ****P < 0.0001 and error bars, SD.

StV-Oriented Transcriptomic Analysis in 299 Mice.

Given their StV defect, 299 mice are an appealing model to identify candidate genes for StV development and maintenance. We chose two time points, E18 and P7, for a preliminary screen of StV-associated genes. E18 is considered the start of StV differentiation in cochlea, in which we expect that transcriptomic analysis may reveal the factors regulating StV development. At E18, we identified 13 differentially expressed genes (DEGs) in 299 mice, as compared to WT mice, including 10 down-regulated and 3 up-regulated genes (SI Appendix, Fig. S9 A and B). However, we observed no statistically significant enrichment in gene ontology (GO) terms among these genes (SI Appendix, Fig. S9C). Interestingly, we found that the transcription factor Esrrb (7) is down-regulated in 299 mouse inner ears at E18. Given the similar inner ear morphology of WT and 299 mice at E18 (SI Appendix, Fig. S4A), we next performed RNA sequencing (RNA-seq) on organ of Corti isolated from WT and 299 mice at P7. We identified a total of 53 DEGs, including 50 down-regulated and 3 up-regulated genes. However, we observed no significant enrichment in GO terms among these genes (SI Appendix, Fig. S10 A–C). As the StV is completely absent in 299 mice at P7, our RNA-seq analysis may be identifying genes involved in StV function and maintenance. Among these 53 DEGs, we observed genes known to regulate StV functions, such as Bsnd, Kcne1, Trpm1, Clcnka, and Kcnq1. Again, Esrrb was found to be significantly reduced in P7 299 LWs. Furthermore, Meis1 is an extremely interesting gene, as it is a transcription factor not previously associated with StV development (SI Appendix, Fig. S10A, red). These RNA-seq data are an example of the immediate applications for 299 mice as an animal model for studying StV development and function as it relates to deafness.

Discussion

In our study, we characterize 299, a genetically distinct hereditary deafness mouse model with an absent StV because of an intergenic insertion of an L1 retrotransposon in chr11: 17,614,267.

The mouse StV begins to develop from the late embryonic stage and is maintained throughout life (20). The 299 mice display absent StV and abnormal scala media development, a unique disease phenotype not previously reported in the literature. The disruption of some StV genes induces a StV defect, albeit with different phenotypic manifestations, including phenotypes in which the StV becomes thinned. Genes related to these phenotypes include Kcnq1 and Kcne1 in MCs (39–41), as well as Mitf and Kcnj10 in ICs (8, 42). However, defects in these genes induce a narrower StV with the deficiency of either K⁺ flux or EP generation (42–44). Nkcc1 mutant mice display the most severe loss of StV (30–32) and Esrrb-knockout (KO) mice display a complete absence of scala media space and a collapse of the RM on to the organ of Corti (7). Interestingly, Pendrin (Slc25a6), which is expressed mainly in outer sulcus cells, seems to have an opposite effect, with Pendrin-KO mice having expended scala media volume (45). Unlike any reported mouse model with StV deficiency, the StV in 299 mice stopped developing as early as P0. Furthermore, 299 mice display a reduced but stably maintained scala media space up until at least at P21, the oldest age we examined (Fig. 3A and SI Appendix, Fig. S4 A–E). Together, compared with a complete StV loss in 299 mice, the thinner StV induced by the disruption of functional genes results in more severe consequences in maintaining scala media space.

The observed StV defect in 299 mice is likely not the result of vasculature deterioration. Instead, vasculature dysregulation is likely a consequence of StV loss. Studies have shown that the mouse StV does not develop until the first week after birth, in which the LW gradually forms the StV and SL with a clear cellular boundary. Meanwhile, the capillaries of the StV and SL are both derived from the spiral cochlear artery, so the blood supplies are theoretically the same (20). In 299 mice, the capillaries in the StV are absent, but SL capillaries are maintained, albeit in a disordered pattern (Fig. 3C and SI Appendix, Fig. S3B). These results suggest that the absence of StV is linked to capillary abnormalities in the LW.

Our findings also suggest a relationship between fluid homeostasis and hair–cell function. In P7 299 mice, MET function is likely enhanced because of the absence of the StV, which in turn is likely due to reduced EP and K⁺ (26). We hypothesize that OHCs then increase the stereocilium number to compensate their MET responsivity in response to the changes in extraciliary fluid conditions. These hypotheses warrant further studies. Interestingly, OHCs do not change their graded MET responses either spatially or temporally, implying that frequency-dependent MET function is likely an intrinsic property of OHCs. In 299 mice with severely disrupted EP and K⁺, basal and middle coil hair cells begin to die at P12 and show extensive loss at 6 wk (19). DPOAE signals are absent at all tested frequencies in P21 299 mice (Fig. 1E), while the apical hair cells still exist (Fig. 2B). These findings suggest that DPOAE depends not only on OHCs but also on the EP. We find that vestibular structures and functions are not affected. However, this may be expected, as endolymphatic factors differ between the cochleae and the vestibules, with roughly +90 mV endolymph potential in the cochleae but less than +10 mV in the vestibular endolymph (4, 46). Considering the different cell layer arrangement in vestibule and StV (47), the factors that control StV differentiation may not be essential for vestibule differentiation.

The deafness-causing mutation in 299 mice arose from a spontaneous L1 retrotransposon insertion and not from a small-scale mutation, as expected from ENU mutagenesis. Our NGS results identified a Ppp3r1 T > C mutation in 299 mice; however, this was not the causative mutation, as shown by the audiometry and histochemical staining of Ppp3r1-KI mice. While short-read NGS is well powered to identify SNP/InDel mutations, there are obvious weaknesses in identifying structural variants, such as large (>200 bp) insertions, deletions, and inversions (34, 48). We indeed observed several gaps when NGS short reads were mapped to repetitive genome sequences (SI Appendix, Fig. S11). L1 has ∼3,000 full-length copies in the mouse genome (49), and sequenced L1 reads have the potential to map to any L1 locus in the 299 genome, including the identified chr11 insertion site. Despite this challenge, we were able to identify a nearly full-length L1-INS on chr11 using PacBio sequencing. PacBio long-read sequencing (∼20 to 35 kb) filled in mapping gaps that were absent in our NGS data (SI Appendix, Fig. S11B). From this, we suggest that whole-genome long-subread sequencing will be useful for identifying pathogenic structural variants in known and yet-to-be identified diseases.

The chr11 L1 retrotransposon insertion affects StV development and function in 299 mice; however, we have yet to identify the effect that our intergenic mutation L1-INS has on genomic regulation. Previous work has revealed that L1-INS in exons or introns can down-regulate the expression of the inserted genes (12). However, there are no annotated genes surrounding the L1-INS, and the nearest coding genes are located >20 kb from the L1-INS. Our RNA-seq results identified dozens of DEGs, many of which warrant further and more detailed studies into how they may be regulated by the L1-INS and for their roles in StV development. Intriguingly, Meis1 is roughly 1.2-Mb downstream of the L1-INS and expressed in the chicken tegmentum vasculosum, which is equivalent to the mammalian StV (50). Gene regulatory sequences usually accompany open chromatin regions (51, 52). It will be useful to map open chromatin regions in the inner ear using assay for transposase-accessible chromatin using sequencing at E18 stage (51). This and similar assays in 299 mice may reveal different regulatory marks and elements (including enhancers, repressors, or promoters) at actively transcribed genes (53). L1-INSs have also been shown to act as a cis regulatory element, affecting surrounding genes (54). We also speculate that multiplex chromatin structural regulation, such as changes in topologically associated domains, could disrupt nearby genes (55, 56). However, these hypotheses require further investigation.

In conclusion, we have characterized a deafness mouse model that is suitable for studying StV development and function. The 299 mice are also an appealing model for studying L1 retrotransposition and its role in cis and trans genomic regulation, as well as how these gene regulation changes contribute to disease phenotypes. Lastly, 299 mice provide a model to identify candidate genes for StV development and maintenance, as suggested by our preliminary transcriptomic analysis and the identification of Meis1 as a potential StV development gene.

Materials and Methods

Mouse Strains and Animal Care.

In this study, the 299 mouse line was a gift from Ulrich Mueller at Johns Hopkins University, Baltimore, MD. The 299 mice were generated from an ENU-induced forward genetic screen project that was performed with funding from NIDCD (National Institute on Deafness and Other Communication Disorders) Grant DC014713 and DC007704 to Ulrich Mueller (when at the Scripps Research Institute). The Ppp3r1-KI mice and 299KO mice were generated in Laboratory Animal Research Center at Tsinghua University. The experimental procedures on mice were approved by the Institutional Animal Care and Use Committee of Tsinghua University.

Audiometry Experiments.

ABR recordings were performed to evaluate hearing threshold, as described previously (57). Briefly, 3 to 4 wk mice were anesthetized by intraperitoneal (i.p.) injection pentobarbitone (with 0.5% pentobarbital sodium at the dosage of 0.2 mL/10 g body weight) and transferred into a sound attenuation chamber. An EC1 speaker was placed inside the external ear canal, and electrodes were inserted under the skin at the vertex and pinna, while a ground was inserted under the skin near the hind leg. Click and pure-tone stimuli were applied starting at an intensity of 90 dB SPL and decreased to 10 dB SPL at 10 dB steps. All the recordings were performed with a TDT system (Tucker-Davis Technologies RZ6). Pure-tone stimuli were applied at 4, 8, 12, 16, and 32 kHz. Click and pure-tone ABR thresholds were analyzed for both ears.

DPOAEs were also performed with the TDT system. Mice were anesthetized and transferred into a sound attenuation chamber, as that used for ABR tests. The stimuli applied in DPOAE test were two related pure tones (f1 and f2, f1 < f2) from separate sound sources with a frequency ration. The stimuli were applied starting at an intensity of 80 dB SPL and decreased to 10 dB SPL at 10-dB steps. A microphone was enclosed with the speaker tubes in the ear to detect signals present in the ear canal. The intensity of 2f1 to 2f2 frequency signal was filtered to evaluate the thresholds of DPOAE. DPOAE stimuli were applied at 4, 8, 12, 16, and 32 kHz.

Facial nerve CAP recordings were conducted, as previously described (15). The 1-mo-old mice were anesthetized, and the facial nerve was exposed from the ventrolateral side of the cochlea. The facial nerve was separated to expose the facial nerve canal. A silver wire was used as the recording electrode that was inserted into the facial nerve canal. The reference electrode and ground electrode were inserted subcutaneously at the pinna and the groin, respectively. The data were collected by BioSigRZ 6 system (Tucker-Davis Technologies). Click stimuli with intensities from 90 to 10 dB SPL at 10-dB step were generated by a TDT MF1 speaker. For better signal noise ratio, 512 responses were averaged at each stimulus level.

Rotarod Test.

Before testing, animals were trained on a rotarod (47650, Ugo Basile) for nine times in 3 d. The training program began at 10 rpm. Mice were put on the rod for 2 to 3 min, and then, the speed increased from 10 to 300 rpm in 80 s. There was a 30- to 45-min interval between each trail. Each animal ran twice per day for three consecutive days. The fall time of each trail was recorded. The total running time was 300 s. All the recorded data were used for data presentation.

Whole-Mount Immunostaining.

Mice at the indicated ages and genotypes were euthanized. Then, the inner ear was dissected from temporal bone and fixed in 4% paraformaldehyde (PFA) (Leagene, DF0135) at room temperature (RT) for 1 h or overnight at 4 °C. Inner ear from mice over 7 d postnatal were decalcified in decalcifying solution (Solarbio, E1171) at RT for 24 h with slowly shaking. For OHCs observation, the basilar membrane was dissected in phosphate-buffered saline (PBS) and blocked in 4% bovine serum albumin (BSA) and 0.5% PBST (PBS + 0.5% Triton X-100 [T8787, Sigma-Aldrich]) at RT for 2 h. Samples were incubated with the Myo7a antibody (1:1,000, Proteus Biosciences Inc. 25–6790) in 4% BSA and 0.5% PBST overnight at 4 °C and washed with PBS three times at RT. After that, samples were incubated with the secondary antibody, including anti-rabbit Alexa Fluor 647 (1:1,000, Life Technology, A21244), Alexa Fluor 488 phalloidin (1:1,000, Life Technology, A12379), and DAPI in 1% BSA and 0.5% PBST at RT for 2 h. Then samples were washed with PBS three times at RT and mounted using ProLong Gold Antifade Mountant (Life Technology, P36930). The OHCs were observed using a Nikon A1R HD25.

For capillary observation, LW was dissected from inner ear in PBS and blocked in 4% BSA and 0.5% PBST at RT for 2 h. Samples were incubated with the CD31 antibody (BD Pharmingen, 550274) in 4% BSA and 0.5% PBST overnight at 4 °C and washed with PBS three times at RT. Samples were then incubated with the secondary antibody, including anti-rat Alexa Fluor 568 (1:1,000, Life Technology, A11077) at RT for 2 h. Then, samples were washed with PBS three times at RT and mounted using ProLong Gold Antifade Mountant (Life Technology, P36930). The morphology of capillaries was observed using a Nikon A1R HD25.

HE staining.

The inner ear of indicated genotype was dissected, fixed, and decalcified, as described in whole-mount immunostaining. After dehydration in 50/70/80/90/95/100/100/100% ethyl alcohol successively, inner ears were incubated twice in xylene for 10 min at RT. Then the xylene in inner ears were replacement by paraffin at 70 °C for 3 h (Leica, EG1150C). Each inner ear was then embedded in paraffin block and sectioned into 5-μm slides using a microtome (Leica, RM2235). The paraffin sections were stained with hematoxylin (Thermo Fisher Scientific, 7211) and eosin (Thermo Fisher Scientific, 7111), according to the manufacturer’s protocol. After HE staining, the samples were imaged with a light microscope (3DHISTECH, Pannoramic SCAN).

SEM.

Basilar membrane and StV from indicated ages and genotypes were fixed, dissected, and dehydrated for SEM, as described before (58). After that, the samples were dried by freeze drying (Hitachi ES-2030) and coated with the gold (Hitachi E-1010). After that, the samples were observed with FEI Quanta 200.

Genotyping.

About 3- to 5-mm length of mouse tail was lysed in alkaline lysis buffer (25 mM NaOH and 0.2 mM EDTA-Na₂, pH 12.0), incubated at 95 °C for 30 min to 1 h, added to the equal volume acid neutralizer (40 mM Tris⋅HCl, pH 5), and mixed well. After that, 2 μL of the solution was used as template for genotyping. Genotyping primers of 299 and Ppp3r1-KI mice were listed in SI Appendix, Table S6. For the genotyping of 299KO mice, about 6- to 8-mm length of mouse tail was used for genomic DNA extraction by the kit (Tiangen, DP304). Genotyping primers were designed outside the cutting site of each sgRNAs. The genotyping primers of 299KO mice were listed in SI Appendix, Table S6. The PCR system and procedure were conducted according to the introduction of KOD One PCR Master Mix (TOYOBO, KMM-101). Targeting PCR products were gel purified and ligated to the pcDNA3.1 plasmid vector and sequenced.

Electrophysiology.

MET currents of hair cells were recorded using the whole-cell voltage clamp, as previously described (58). Briefly, the basilar membrane was dissected from the cochlea of mice with indicated genotypes and ages. The dissection solution contained the following (in millimolar): 141.7 NaCl, 5.36 KCl, 0.1 CaCl₂, 1 MgCl₂, 0.5 MgSO₄, 3.4 L-glutamine, 10 glucose, and 10 H-Hepes (pH 7.4). Then the basilar membrane was transferred into a recording chamber with the recording solution containing the following (in millimolar): 144 NaCl, 0.7 NaH₂PO₄, 5.8 KCl, 1.3 CaCl₂, 0.9 MgCl₂, 5.6 glucose, and 10 H-Hepes (pH 7.4). All experiments were performed at room temperature (20 to 25 °C).

To record EP in scala media of cochlea, 3 to 4 wk mice were anesthetized i.p. with 4% pentobarbital sodium. The left cochlea was exposed with a ventrolateral approach. The bulla near the round window was carefully removed to expose the round window. The sharp glass electrode with a tip of 0.1- to 0.2-μm diameter was pulled by pipette puller (Sutter, P-2000). The electrode was filled with 150 mM KCl and inserted into the scala media through the round window driven by a micromanipulator (Luigs & Neumann, mini 25). The Ag-AgCl reference electrode was placed in the neck muscle. The baseline of recorded potential was offset to zero when the tip of electrode reached the round window before penetrating into the scala media. The data were collected by an electrophysiology amplifier (Axon, 700B) powered with WinWCP software (University of Strathclyde Glasgow).

To record EP and K⁺ concentration in scala media of cochlea simultaneously, the double-barreled capillary glass (World Precision Instruments) was thoroughly cleaned before use. Firstly, soaking in chromic acids followed by thorough rinsing in distilled water and finally drying in an oven. After trimming one barrel by ∼1 cm, the electrodes were pulled by a microelectrode puller (P-97, Sutter). The diameter of the double-barreled electrode tip was 2 to 3 μm. A 100-mL beaker covered with silicon membrane was acted as the silanization chamber. The beaker and the electrode were baked for 30 min at 150 °C in an oven to eliminate all the moisture before the silanization. Then, the long barrel of the electrode was inserted onto the silicon membrane, and after the injection of 20 μL dimethyldichlorosilane, the beaker and the electrode were baked for another 30 min and cooled in a desiccator. Following silanization, the long barrel was filled with 150 mM KCl and about 200-μm-long K⁺ exchanger (Potassium ionophore I, Sigma-Aldrich) was filled at the tip of the long barrel. The reference barrel was filled with 150 mM NaCl. The electrode was calibrated before and after an experiment with the standard solution of (in millimolar) 10 KCl: 190 NaCl, 50 KCl: 150 NaCl, 100 KCl: 100 NaCl, and 200 KCl at 37 °C. The recording results were used to convert measured K⁺ diffusion potentials to ionic activity. The following steps were performed as EP recoding with the single sharp electrode.

High-Throughput Sequencing.

Whole-genome NGS was performed as described before (59). In brief, genomic DNA from 299 heterozygous and homozygous mice was extracted by a genomic extraction kit (Tiangen, DP304). The genomic DNA was broken into small fragments (∼350 bp) randomly. Then, the sequencing library was generated using TruSeq Library Construction Kit and sequenced using the Illumina Hiseq platform. PacBio sequencing was performed as described before (36). In brief, the whole genomic DNA of 299 homozygous mice were extracted, and SMRT sequencing library was generated and sequenced on the PacBio sequel system.

CRISPR/Cas9-Mediated Deletion of L1 Insertion in 299 Mice.

Target sites for CRISPR/Cas9 were designed by CRISPR/Cas9 target online predictor (CCTop) tools (60). To delete the 7 kb L1-INS in 299 mice, two sgRNAs were chosen at either side of the endpoints on the same strand. The sequence of sgRNAs for the 299KO experiment were listed in SI Appendix, Table S7. ssODN was designed to bridge the flanked sequences after deletion of the 7 kb. ssODN was designed 120 bp in length and positioned directly adjacent to the most popular cutting sites of each guide RNA. The sequence of ssODN was listed in SI Appendix, Table S7. The Cas9 RNA and sgRNAs were synthesized as described previously (38). In brief, the T7-Cas9 plasmid was linearized. The double-strand DNA templates for sgRNA production were cloned from sgRNA plasmid vector using Phanta high-fidelity enzyme (Vazyme, P505) and purified by DNA extraction kit (Magen, D2111). The Cas9 RNA and sgRNAs were transcribed and purified in vitro using the HiScribe T7 Quick High Yield RNA Synthesis Kit, according to the instruction manual from New England Biolabs (NEB, E2050S). The zygote injection procedure was conducted by Laboratory Animal Research Center in Tsinghua University, as described previously (38). The injection solution was prepared in ribonuclease (RNase) free water, which had 60 ng/μL Cas9 RNA, 30 ng/μL sgRNA (each), and 100 ng/μL ssODN.

Generation of Ppp3r1-KI Mouse Line.

Target sites for CRISPR/Cas9 were designed by CCTop tools (60). The sequence of sgRNAs for Ppp3r1-KI experiment were listed in SI Appendix, Table S8. The Cas9 RNA and sgRNAs were synthesized, as described, in the generation of 299KO mice. The injection solution was prepared in RNase free water, which had 25 ng/μL Cas9 RNA, 10 ng/μL sgRNA, and 25 ng/μL ssODN.

RNA Sequencing.

Pregnant female mice were used for collecting E18 embryos (five for each genotype), and the inner ears were collected. For P7 pups, the organ of Corti samples were acquired. We extracted the total RNA according to the manual of HiPure Unviersal RNA Mini Kit (Cat. R4130, Magen). The total RNA samples were measured for the concentration and purity by Nanodrop 2000c (Thermo Fisher Scientific) and were quality controlled by agarose gel electrophoresis, respectively. Then, the total RNA samples were sent for messenger RNA sequencing and data analysis (Novogene).

Data Analysis.

Blinded analyses were conducted on histological data. For HE staining samples, only one section in all sections, from one cochlea, shows that most complete modiolus were chosen for analysis and only the third organ of Corti from a cochlea section was used for measurements. To quantify the loss of StV, we calculated the ratio of lengths and areas as shown in Fig. 3B. To quantify numbers of spiral ganglion neurons, we counted cell numbers in a fixed, round-shape area as shown in SI Appendix, Fig. S2A. Data were managed and analyzed with Excel 2016 (Microsoft), Prism 6 (GraphPad Software), and Igor pro 6 (WaveMetrics). All data are shown as mean ± SD, as indicated in the figure legends. We used two-tailed Student’s t test for one-to-one comparison to determine statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). n numbers are indicated in the figures.

Data Availability

All study data are included in the article and/or SI Appendix.