Expression of amyloid-β in mouse cochlear hair cells causes an early-onset auditory defect in high-frequency sound perception

Establishment of Tg mouse lines expressing Aβ42 or its derivative in hair cells of the cochlea

To express Aβ or its truncated or mutated forms in mouse cochlear hair cells, we established Tg mouse lines carrying an expression vector for Aβ, a truncated form of Aβ (Aβ42), or Aβ with a familial AD mutation (Aβ42^Arc, E22G substitution) downstream of the minimal enhancer region of mouse Math1 (Math1^E), which drives gene expression in many types of sensory neurons including cochlear hair cells [24] (Fig. 1A). These constructs were used for expression in mouse cochlear hair cells. The rat prepro-enkephalin signal sequence was fused upstream of the Aβ sequences to ensure their secretion [25]. Quantitative PCR (qPCR) showed that Aβ42 and Aβ42^Arc mRNAs were well expressed in the cochleae of Tg mice compared with non-Tg control mice (Fig. 1B). Enzyme-linked immunosorbent assay (ELISA) revealed the presence of Tris-buffered saline–soluble and –insoluble (formic acid–soluble) forms of an Aβ protein in the cochleae of 4-month-old Tg(Math1^E-Aβ42^Arc)1Lt mice (Fig. 1C).

Aβ42^Arc is found around the hair cells of Tg mouse

We surveyed the cochlear hair cells of Tg(Math1^E-Aβ42^Arc)1Lt mice histochemically and found that the structures of hair cells of Tg-mice were not different from those of non-Tg siblings at age 4 months (Fig. 2A, B). Next, staining with anti-Aβ (6E10) revealed that Aβ proteins were distributed mainly around the cochlear hair cells of 4-month-old Tg(Math1^E-Aβ42^Arc)1Lt mice (Fig. 2C, D). Aggregated Aβ proteins were also found near these hair cells, although they were not as abundant (Fig. 2E, arrows). To identify the precise cellular location of Aβ in the Tg(Math1^E-Aβ42^Arc)1Lt mice, we performed immunoelectron microscopy. Aβ immuno-staining was observed in the endoplasmic reticulum, the stereo cilia, and the plasma membrane of the hair cells (Fig. 2F, G and H). Notably, Aβ was found as clusters comprised of 2–20 particles in the stereo cilia and plasma membrane (Fig. 2G, H, arrowheads). This is similar to previous results that Aβ tends to exist as oligomeric clusters in the central nervous system [26]. Interestingly, Aβ immunostaining was also observed within and on the surface of the surrounding cells, suggesting that Aβ-like protein might be secreted to the cells neighboring the hair cells (Fig. 2H; arrows). These finding are in agreement with previously observed properties of Aβ-related peptides in the central nervous system [25, 27, 28].

Expression of Aβ derivatives in cochlear hair cells impairs the perception of high-frequency sound

To assess the consequence of Aβ42 or Aβ42^Arc expression in cochlear hair cells, we monitored hearing with ABRs. ABRs are electrical action potentials in the cochlea and brain induced by an auditory stimulus and can be detected via electrodes placed on the scalp [29]. We monitored threshold for the first peak of the ABR response (1–2 ms after sound stimulation) because the first peak is generated exclusively by the activity of sensory hair cells and auditory nerves [19]. We first examined the response to a click sound, which is not frequency tunable but reflects the activity in a broad region of the cochlear hair cells, and we compared Tg mice and non-Tg mice that were wild-type littermates of the Tg mice. We found that the sound threshold for the non-Tg, Tg(Math1^E-Aβ42)1Lt, and Tg(Math1^E-Aβ42^Arc)1Lt mice were not significantly different from age 2 to at least 8 months, suggesting that sound perception itself was not affected by the expression of Aβ in the hair cells (Fig. 3A, D). Next, we monitored the response to a high-frequency tone (32 kHz) and found that the thresholds for each of the Tg(Math1^E-Aβ42)1Lt and Tg(Math1^E-Aβ42^Arc)1Lt mice was slightly increased from age 3 to 5 months (Fig. 3B, E, arrows), whereas the mid-range frequency (8 kHz) threshold of the Tg(Math1^E-Aβ42)1Lt and Tg(Math1^E-Aβ42^Arc)1Lt mice was comparable to that of the non-Tg siblings during this period (Fig. 3C, F).

Although there were apparent differences in the ability of Tg mice to hear high-frequency sound, the ABR threshold increased with age for all mouse strains, and after the mice were 6 months old, we could no longer detect a difference between non-Tg and Tg mice.

The confluence of threshold might be a consequence of the genetic background of the Tg mice, which is C57BL/6 (B6). B6 mice have a point mutation in Cadherin 23, which induces an age-related hearing defect that leads to an increased dB threshold after age 6 months (Supplemental Fig. S1) [30, 31]. It has been shown that B6 and 129/SvJ (129) heterozygotes do not suffer from this age-related hearing defect [32]. Thus, we crossed Tg(Math1^E-Aβ42^Arc)1Lt mice with the 129 line and measured ABR of the heterozygotes (Tg(Math1^E-Aβ42^Arc)1Lt/129). We first compared the response to high-frequency (32 kHz) tone sound stimulation at 2 and 4 months of age and found that hearing ability was improved in the B6/129 background; the heterozygotes responded to a 40 dB sound stimulation, whereas the B6 mice no longer responded to this stimulation at age 4 months (Fig. 4A, B, compare to Fig. 3). We found that both non-Tg (B6/129) and Tg(Math1^E-Aβ42^Arc)1Lt/129 mice showed a similar response at age 2 months, whereas the first peak could no longer be detected in Tg(Math1^E-Aβ42^Arc)1Lt/129 mice at 4 months (Fig. 4A, B, arrow).

We then determined threshold levels at age 2 and 4 months and found that the Tg(Math1^E-Aβ42^Arc)1Lt/129 mice showed a greatly increased threshold relative to the non-Tg (B6/12 9) mice at age 4 months (Fig. 4C).

Next, we monitored changes in threshold level from age 2 to 8 months in Tg(Math1^E-Aβ42^Arc)1Lt/129 mice and confirmed that the mice showed an increasing threshold for high-frequency (32 kHz) sound stimulation from age 4 to at least 8 months (Fig. 4D). The hearing impairment of Tg(Math1^E-Aβ42^Arc)1Lt/129 mice seemed to be exclusively in high-frequency sound perception; the threshold of the Tg(Math1^E-Aβ42^Arc)1Lt/129 mice did not change for perception of lower-frequency sound (16 and 8 kHz) or click sound (Fig. 4E–G). These results suggeste d that Aβ42^Arc expression in the hair cells exclusively affected the ability of the mice to respond to high-frequency sound (>32 kHz). To confirm that the defect in Tg(Math1^E-Aβ42^Arc)1Lt/129 mice was exclusively in high-frequency sound perception, we analyzed the threshold response to a 40 kHz tone at age 8 months and found that the threshold was still significantly higher for Tg mice (Supplemental Fig. S2).

Expression of Aβ derivatives in the hair cells causes loss of hair cells in the basal region of cochlea

High-frequency sound is perceived by the hair cells in the basal region of the cochlea [19]. Previous studies have suggested that the hair cells in the basal region of cochleae are lost during aging [4]. To elucidate the mechanism behind the defect in the ability to perceive high-frequency sound at age 8 months, we stained surface preparations of the organ of Corti derived from 8-month-old non-Tg and Tg(Math1^E-Aβ42^Arc)1Lt/129 mice with Alexa Fluor 488–phalloidin and DAPI and visualized the hair cells (Fig. 5A). Non-Tg epithelia had one row of inner hair cells and three rows of outer hair cells at the basal, mid, and apical levels of the cochlear duct [1, 2]. We compared the hair cells in each region of the cochlea and found that the number of hair cells appeared to be comparable between non-Tg and Tg mice in the apical and middle regions of cochlea (the region responsible for perceiving frequencies <20 kHz) (Fig. 5A–C, F). However, the number of hair cells in the basal region of the cochlea (region responsible for perceiving frequencies >20 kHz) was dramatically reduced in the Tg mice (Fig. 5D–F). These results suggested that expression of Aβ42^Arc specifically affects the survival of hair cells in the basal region of cochlea. This is correlated with the fact that neuronal cell death in the central nervous system is typical of AD pathology [33].

Expression of the human microtubule-associated protein tau (MAPT) in cochlear hair cells exacerbates the hearing defects caused by Aβ42^Arc

MAPT is thought to be another major player in the development of AD [34]. Many studies indicate that there are synergistic effects between Aβ and tau in the neuronal toxicity of AD in the brain [35, 36]. To elucidate if Aβ42^Arc-dependent high-frequency hearing defect reflects other aspects of AD in the brain, we looked for genetic interaction between Aβ42^Arc and tau in terms of hearing defects in the cochlear hair cells. We established another Tg mouse line, Tg(Math^E-MAPT)1Lt, that expressed human tau (2N4R) in cochlear hair cells, and we monitored auditory function of these mice over time (Supplemental Fig. 3A, B). No significant defects in hearing were seen in these mice when exposed to sound frequencies of 32 or 8 kHz or to click sounds (Fig. 6A–C). To determine whether there is interaction between Aβ42^Arc and tau, we established a double Tg mouse line (Tau/Aβ42^Arc) harboring both Tg(Math1^E-Aβ42^Arc)1Lt and Tg(Math^E-MAPT)1Lt. We found that these mice had severe hearing defects; the ABR threshold was significantly increased at age 2 and 3 months for both the 32 and 8 kHz tones (Fig. 6D, E). However, we did not detect an increase in the threshold for the click sound, suggesting that the hearing defect was only in the middle- to high-frequency sound perception (>8 kHz; Fig. 6F). The significant shift in the onset of the hearing defect from age 4 months in the Aβ42^Arc mice to age 2 months in the tau/Aβ42^Arc double Tg-mice strongly suggested that there is a synergistic interaction between Aβ and tau in this system.

The sensory system could be a useful tool to study age-related neurodegenerative disorders

It has been suggested that sensory impairments are involved in neurodegenerative disorders because olfactory defects appear to correlate with AD, mild cognitive impairment, Parkinson's disease, and Huntington's disease [15–18]. The auditory systems are also involved in the regulation of postural stability in AD patients [37], and there seems to be a correlation between deafness and AD-related dementia [38]. In mice, it has been shown that retinal expression of Aβ in APPswe/PS1DeltaE9, a mouse AD model, causes retinal dysfunction, suggesting that sensory decline with age-related neurodegenerative disorders might be evolutionally conserved [39]. In support of this notion, many studies have shown that compound eyes of Drosophila melanogaster can serve as a model for age-related neurodegenerative disorders such as AD, Parkinson's disease, and Huntington's disease [40–42].

Overproduction of the causative factors for those disorders in photoreceptor cells results in impaired survival of those cells. Notably, many of the molecules related to the pathological properties of these diseases in the brain were identified by genetic analysis using this system [43–45]. These results suggest that sensory system models for age-related neurodegenerative disorders might reflect certain aspects of neuronal defects in the brain.

In this study, we established a new AD model system that takes advantage of the mouse auditory system. We produced new Tg mice overexpressing Aβ or Aβ-related molecules in cochlear hair cells. The Tg mice showed auditory defects in high-frequency sound perception and synergistic interaction between Aβ and tau in terms of hearing impairment and hair cell loss (Figs. 4, 5, 6). Given that genetic interaction between Aβ and tau and loss of neuronal cells are typical pathological futures of AD [33, 34], we believe that our system reflects many, if not all, aspects of AD progression.

Relationship between Aβ-induced hearing defects and ARHL

During aging, the survival of auditory hair cells decreases, especially in the basal region of the cochlea, and these cells are thought to be very sensitive to stressors such as excessive noise exposure [4]. This vulnerability of hair cells is likely the basis of the development of ARHL, as many animals with ARHL-related mutations show auditory defects after several hours of exposure to extreme noise [5]. We examined whether vulnerability of hair cells is involved in the hearing defects of Tg(Math1^E-Aβ42^Arc)1Lt mice by investigating if exposure to white noise at 100 dB for 2 h would affect the hearing ability of 3-month-old mice. One week after exposure, we measured the ABR threshold, however, no significant defects in hearing were seen in Tg(Math1^E-Aβ42^Arc)1Lt mice when exposed to sound frequencies of 32, 8 kHz or to click sounds (data not shown). Therefore, we think that auditory defects caused by Aβ differ from those associated with ARHL.

Tg(Math1^E-Aβ42^Arc)1Lt mice are a potential new assay system for the study of AD

To investigate the toxic effects of Aβ during AD development, several different types of mouse models for AD have been established, many of which express human APP in the brain [46]. These APP-expressing mice are very useful as an assay system for senile plaque formation in the brain and for learning/memory impairment [46]. However, these APP-expressing mice have several disadvantages as a model system for the development of drugs against AD. First, it takes approximately one year to obtain the robust phenotype caused by Aβ toxicity. Second, the progression of disease in these systems is based on behavioral symptoms, which are assessed by a behavioral test such as the water maze or Barnes test [46]. Thus, these systems are very useful for measuring higher-order brain functions, such as learning and memory, but they are not well suited for quantitative analysis [46]. Third, APP-expressing mice do not show neuronal cell loss, which is a typical pathological feature found in AD patients [33, 34].

Herein, we describe a new Tg system to explore AD based on Tg(Math1^E-Aβ42^Arc)1Lt mice. Our Tg-mouse system can easily monitor the effect of Aβ toxicities using electrophysiological methods. In other words, we can assess the impact of Aβ in live animals quantitatively over time. In particular, this system can also detect cellular defects like hair cell loss when Aβ is expressed. Therefore, we expect that this system will be a powerful tool when it is complementary used with traditional AD mouse models for evaluating the effects of Aβ toxicity and developing therapeutic drugs against AD.

Animals

Animal care and treatment were performed in accordance with institutional guidelines following approval by the Animal Care and Use Committee of National Center for Geriatrics and Gerontology (NCGG) (Obu, Japan). Before treatments, animals were anesthetized with isoflurane (Wako).

Mice were obtained from breeding colonies of Tg(Math1^E-Aβ42)1Lt, Tg(Math1^E-Aβ42^Arc)1Lt, and Tg(Math^E-MAPT)1Lt, which were heterozygous for the transgene, and were maintained at the Animal Center of NCGG. The Tg mice were maintained in a mixed C57BL/6 background and back-crossed to generate Tg(Math1^E-Aβ42)1Lt/B6, Tg(Math1^E-Aβ42^Arc)1Lt/B6 or Tg(Math^E-MAPT)1Lt/B6. We refer to the resulting lines as Tg(Math1^E-Aβ42)1Lt, Tg(Math1^E-Aβ42^Arc)1Lt, and Tg(Math^E-MAPT)1Lt, respectively. Wild-type littermates of the Tg mice served as controls (non-Tg mice). The Tg(Math1^E-Aβ42^Arc)1Lt mice were also crossed with 129 to generate Tg(Math1^E-Aβ42^Arc)1Lt/129.

Vector construction

We cloned the mouse Math1 minimal enhancer (Math1^E) sequence into a pBluescript plasmid (Stratagene) containing the cytomegalovirus promoter and IRES-EGFP::L10-β-globin heavy chain-poly(A) sequence [24]. Aβ42 and Aβ42^Arc constructs containing an upstream rat prepro-enkephalin signaling sequence, which had previously been constructed from cDNA of APP by PCR, or MAPT were placed downstream of the cytomegalovirus promoter [47, 48]. The mouse Math1^E was kindly provided by Dr. J. E. Johnson (University of Texas Southwestern Medical Center, Dallas, TX).

Real-time qPCR

The qPCR was performed as described [49]. Briefly, RNA was extracted from cochleae using Isogen (Wako) and the RNeasy Mini kit (Qiagen), and the different cDNAs were individually synthesized with Primescript RT reagent (Takara). The mRNA expression levels were quantified using a Thermal Cycler Dice Real-Time System and SYBR Premix Ex Taq (Takara). Data were normalized to β-actin mRNA expression level. Myo7a, unconventional myosins playing an important role in morphogenesis and organization of cochlear hair cell bundles, was used as a positive control. The thermal cycling parameters were 40 cycles at 95°C for 10 s and 60°C for 30 s.

Oligonucleotides for real-time PCR analysis were as follows.

• β-actin:

  • sense, 5′-CTAAGGCCAACCGTGAAAAG-3′

  • antisense, 5′-ACCAGAGGCATACAGGGACA-3′

• Aβ:

  • sense, 5′-TCCGACATGACTCAGGATATGA-3′

  • antisense, 5′-CCCACCATGAGTCCAATGA-3′

• myo7a:

  • sense, 5′-GCCATTGCTGACAACTGCTA-3′

  • antisense, 5′-TTGTGCTCTCTGTCTTGCCA-3′

ELISA

Soluble and insoluble (formic acid extractable) fractions were prepared from extracts of cochleae from Tg(Math1^E-Aβ42^Arc)1Lt and non-Tg mice. Briefly, frozen cochleae were homogenized with a metal pestle in Tris-buffered saline containing protease inhibitor cocktail (Complete Mini, Roche). After centrifugation at 100,000 × g for 1 h at 4°C, the soluble fraction was subjected to ELISA using the High Sensitivity Human β-Amyloid ELISA kit (Wako) to detect Aβ-like protein. The pellet was treated with 2% (w/v) sodium dodecyl sulfate and then 70% (v/v) formic acid and centrifuged at 100,000 × g for 1 h at 4°C. The supernatant, which represented the insoluble fraction, was subjected to ELISA.

Histochemistry

Cochleae were isolated from Tg(Math1^E-Aβ42^Arc)1Lt and non-Tg mice at age 4 months and fixed in paraffin. Paraffin sections were stained with hematoxylin and eosin or immunostained using Aβ antibody 6E10 (BioLegend).

Epon-embedded tissue sections were prepared as described [47]. Briefly, dissected cochleae were fixed in 2.5% (v/v) glutaraldehyde and decalcified for 3 days at 4°C in 5 mM EDTA, 0.1 M sodium phosphate (pH 7.4). Tissues were then fixed with 1% (w/v) OsO₄ (Wako) in PBS (pH 7.4) for 1 h at room temperature. Tissues were then washed with PBS for 2 h with the wash solution changed every 15 min and then dehydrated as follows: once for 15 min in 50% (v/v) ethanol, twice for 30 min in 70% (v/v) ethanol, and once for 30 min in 85% (v/v) ethanol. After treatment with propylene oxide (Nisshin EM Corporation.), tissues were embedded in Epon 812 resin (TAAB Laboratories Equipment.) and incubated at 60±C for 48 h.

Phalloidin staining was performed by the standard protocol [1]. Briefly, dissected cochleae were fixed overnight with 4% paraformaldehyde in PBS. After cochleae were decalcified for 3 days at 4°C in 5 mM EDTA, 0.1 M sodium phosphate, they were incubated with phalloidin antibody coupled to Alexa Fluor 488 (Thermo). Dissected cochleae were mounted with Vectashield Mounting Medium that contained DAPI (Vector laboratories).

ABR measurement

ABRs were recorded in anesthetized mice. Electro-brainstem responses to the click (100 ms, 0–90 dB sound pressure level) and pure tonal sounds (8–32 kHz in half-octave steps, 20–100 dB sound pressure level in 5-dB steps, 3 ms, 1-ms cosine-squared rise-fall envelope) were recorded with subdermal silver wire electrodes at the ear (positive, active), the vertex (negative, reference), and the back of the mice (ground). ABR recordings were analyzed for consecutive amplitude deflections (waves). The first peak (wave I; latency, 1.2–1.9 ms) was monitored, and the hearing threshold for a given frequency or click stimulus was calculated for the lowest sound pressure leading to reliable ABR signals. Each ABR threshold was obtained by comparing the amplitude levels 5 dB above the hearing threshold for each group. Recordings were made for 12.5 ms with stimulus presentations of alternating polarity to eliminate electrical artifacts. In each case, stimulus presentation was at 0 ms. Signals were amplified (1000-fold, 40–60 dB), band pass filtered (0.2–5 kHz Butterworth, −20 dB/decade), averaged across 64–256 repetitions (dependent on the signal-to-noise ratio) at each sound pressure (usually 0–100 dB sound pressure level in steps of 5 dB), and recorded at 40 kHz sample frequency. Stimuli were delivered to the ear in a calibrated open system by a loudspeaker (model ES-1, TDT) placed 5 mm lateral to the pinna by an ear coupler. Sound pressure was calibrated online prior to each measurement with a microphone (NA42S, RION) placed under the scalp near the ear.

Mice were exposed to inhaled isoflurane anesthesia during ABR recording using a vaporizer.

Data analysis

For statistical analyses, we used the Microsoft Excel Macro program (Microsoft) and JMP (SAS Institute).