Orphan Glutamate Receptor δ1 Subunit Required for High-Frequency Hearing

ABSTRACT The function of the orphan glutamate receptor delta subunits (GluRδ1 and GluRδ2) remains unclear. GluRδ2 is expressed exclusively in the Purkinje cells of the cerebellum, and GluRδ1 is prominently expressed in inner ear hair cells and neurons of the hippocampus. We found that mice lacking the GluRδ1 protein displayed significant cochlear threshold shifts for frequencies of >16 kHz. These deficits correlated with a substantial loss of type IV spiral ligament fibrocytes and a significant reduction of endolymphatic potential in high-frequency cochlear regions. Vulnerability to acoustic injury was significantly enhanced; however, the efferent innervation of hair cells and the classic efferent inhibition of outer hair cells were unaffected. Hippocampal and vestibular morphology and function were normal. Our findings show that the orphan GluRδ1 plays an essential role in high-frequency hearing and ionic homeostasis in the basal cochlea, and the locus encoding GluRδ1 represents a candidate gene for congenital or acquired high-frequency hearing loss in humans.

Ionotropic glutamate receptors include three major families, N-methyl-D-aspartate (NMDA), kainate, and ␣-amino-3-hydroxy-5-methyl-4-isoxazole-4-propionic acid (AMPA) receptors, and a fourth orphan family of delta receptors (GluR␦1 and GluR␦2). GluR␦1 and GluR␦2 share 56% amino acid identity with each other but only 17 to 28% identity with other ionotropic glutamate receptors (25,45). Neither GluR␦1 nor GluR␦2 can be activated by AMPA, kainate, NMDA, glutamate, or any other ligands when expressed alone or in combination with other subunits in heterologous expression systems (46). Sequence analysis suggests that both GluR␦1 and GluR␦2 are more homologous to non-NMDA receptors; an analysis of GluR␦2 with the Lurcher mutation in the third transmembrane domain suggests that GluR␦2 functions as an AMPA-like receptor (21,44,48). It is expressed exclusively in Purkinje cells of the cerebellum. Targeted disruption of GluR␦2 causes motor coordination impairment, Purkinje cell maturation, and long-term depression of synaptic transmission (20). Subsequently, it was found that the appropriate transport of GluR␦2 to the Purkinje cell surface is required for the function of the receptor in synaptic transmission (15,46). Recently, it has been suggested that GluR␦2 is the receptor for cerebellin 1, a glycoprotein of the C1q and tumor necrosis factor family that is secreted from cerebellar granule cells (16).
In contrast to GluR␦2, GluR␦1 is expressed in many areas in the developing central nervous system, including the hippocampus and the caudate putamen, but is absent in the cerebellum in guinea pigs, rats, and mice (24)(25)(26). In the adult, GluR␦1 is expressed at high levels in hippocampal neurons (24)(25)(26), cochlear inner hair cells (IHCs), and spiral ganglia and their satellite cells as well as vestibular hair cells and vestibular ganglia in guinea pigs and rats (36). GluR␦1 is also weakly expressed in Claudius cells in the basal cochlear turns and in vestibular supporting cells (36). In IHCs, GluR␦1 immunostaining appears over the entire cell surface rather than localized to the synaptic sites at the base of the cell (36). Despite these expression patterns, no functional role of GluR␦1 in vivo has been reported.
To investigate its role, we created and characterized a null allele of GluR␦1 in mice. The GluR␦1 Ϫ/Ϫ mice displayed a significant auditory phenotype, demonstrating that GluR␦1 is required for high-frequency hearing and suggesting that it has a role in cochlear ion homeostasis. Function in other cells expressing GluR␦1, such as vestibular hair cells, vestibular ganglia, and hippocampal neurons, was not significantly affected in GluR␦1 Ϫ/Ϫ mice. The locus encoding GluR␦1 thus represents a candidate gene for congenital or acquired highfrequency hearing loss in humans. Table 1 summarizes the procedures used in these studies as well as the age and number of mice used for each procedure. Details for each procedure follow.

MATERIALS AND METHODS
Construction of the GluR␦1 targeting vector and generation of GluR␦1 mutant mice. We screened a bacterial artificial chromosome library (Research Genetics) containing mouse 129/Sv genomic DNA and obtained overlapping clones with average sizes of 150 kb. A 7-kb GluR␦1 genomic DNA fragment containing exon 11 (transmembrane domains 1 and 2 [TM1 and TM2]) and another 10-kb fragment containing exon 12 (TM3) were isolated, restriction mapped, and sequenced. TL-1 embryonic stem (ES) cells derived from the 129/SvEv strain were electroporated with linearized targeting vector. DNA from the ES cell line was digested with SpeI and analyzed by Southern blotting (Fig.  1). One homologously recombined targeted cell line was obtained, and subsequent germ line transmission was achieved. We developed a PCR genotyping assay (30 cycles) with a pair of primers from the deleted region of the GluR␦1 gene (5Ј GCAAGCGCTACATGGACTAC 3Ј and 5Ј GGCACTGTGCAGGG TGGCAG 3Ј) and a pair of primers from the targeting vector (5Ј CCTGAAT GAACTGCAGGACG 3Ј and 5Ј CGCTATGTCCTGATAGCGATC 3Ј). All mice analyzed were from a mixed background of 129/SvEv and C57BL/6 in the F 2 to F 6 generations. All mouse use protocols were approved by institutional animal care and use committees.
Immunostaining and histologic analysis. For the evaluation of molecular and morphological changes in GluR␦1 Ϫ/Ϫ mice, mice were anesthetized with Avertin (500 mg/kg of body weight) or ketamine and xylazine (0.72/0.46 mg/30 g of body weight), followed by intracardial perfusions of 0.1 M phosphate-buffered saline and subsequently 4% paraformaldehyde solution in 0.1 M phosphate buffer (pH 7.3). Cochleas were postfixed overnight and then decalcified in EDTA for 1 to 3 days. For whole-mount immunolabeling, cochleas were dissected, permeabilized with 1.0% Triton X-100 for 10 min, and incubated overnight in primary antibodies. The next day, the samples were placed in biotin-labeled secondary antibodies, a complex consisting of avidin, biotin, and horseradish peroxidase (ABC kit; Vector Laboratory), and then incubated in peroxidase substrate. For immunostaining of sections, decalcified cochleas were embedded in paraffin and sectioned into 12-m thicknesses. Slides were deparaffinized. Nonspecific binding of secondary antibody was blocked by incubation with 10% goat serum in phosphate-buffered saline for 30 min at room temperature. Samples were then incubated in primary and secondary antibodies as described above. Samples were observed under a microscope (Olympus BX60). The primary antibodies used were Chemicon AB1514 for GluR␦1/2, Abnova H00002894-A01, specific for GluR␦1 (for hippocampal immunostaining), Santa Cruz SC22926 for Kir3.1, Sigma P6610 for Kv4.1, Santa Cruz SC16053 for the Na ϩ -K ϩ ATPase ␤1 subunit (Atp1b1), Santa Cruz SC21547 for the Na ϩ -K ϩ -2Cl Ϫ cotransporter (Nkcc1), Chemicon MAB329 for synaptophysin, Sigma V5387 for the vesicular acetylcholine transporter (VAT), Chemicon AB197 for the calcitonin gene-related peptide (CGRP), and Chemicon AB5811P for SNAP25. For the anti-GluR␦1/2 antibody from Chemicon, lots produced before 2002 worked well in our immunostaining and Western blot analyses, but recently produced lots failed in immunostaining. In our hands, the various GluR␦1-specific antibodies (Abnova H00002894-A01 and Abnova H00002894-M01; kindly provided by R. Wenthold) did not result in immunostaining signals that were consistent and different between GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ cochlear sections despite numerous attempts with a variety of conditions, including antigen retrieval.
For the immunostaining of SNAP25, synaptophysin, CGRP, or VAT, each dissected cochlear piece was measured by computerized planimetry and the cochlear location was converted to the frequency that is normally processed at that location (8). To quantify immunopositive terminals, outlines were traced via a drawing tube using high-numerical-aperture objective lenses (total magnification, ϫ2,000). During tracing, fine focus was continually adjusted to optimize  11 and 12 is replaced with the PGK-Neo-pA (Neo) cassette in the targeting construct. A 2.9-kb fragment (short arm) and a 4.7-kb fragment (long arm) were used. A new SpeI (S) site in the Neo cassette resulted in a shorter band after homologous recombination on the Southern blot when the external probe from exon 10 was used. H, HindIII; Hp, HpaI; P, PstI. (B) Southern blot analysis of genomic DNA isolated from GluR␦1 ϩ/ϩ (ϩ/ϩ), GluR␦1 ϩ/Ϫ (ϩ/Ϫ), and GluR␦1 Ϫ/Ϫ (Ϫ/Ϫ) mice. SpeI-digested tail DNA was hybridized with the external probe (exon 10). The probe detected 15-kb wild-type (WT) and 6-kb homologous recombined (HR) bands. (C) PCR genotyping assay using a pair of primers (from the deleted region of TM1 to TM3) that amplify a 220-bp band and another pair of primers (from the targeting vector) that amplify a 500-bp band (see Materials and Methods). Note that there is no 220-bp band in the homozygous mouse.
VOL. 27,2007 GluR␦1 REQUIRED FOR HIGH-FREQUENCY HEARING 4501 imaging of each terminal cluster. Traces were digitized, and areas were computed using NIH Image software. For the outer hair cell (OHC) area, all immunopositive terminals were traced and values from each row were averaged within bins corresponding to 100 m of cochlear length. For an assessment of histopathology, animals were anesthetized, followed by intracardial perfusion with 2.5% glutaraldehyde and 1.5% paraformaldehyde in phosphate buffer. Temporal bones were extracted, and round and oval windows opened for intralabyrinthine perfusion of fixative. Cochleas were then osmicated (1% OsO 4 in dH 2 O), decalcified (0.1 M EDTA with 0.4% glutaraldehyde), dehydrated in ethanols and propylene oxide, embedded in Araldite resins, and sectioned at 40 m on a Historange with a carbide steel knife. Sections were mounted on slides and coverslipped.
Laser capture microdissection of cochlear sections and reverse transcriptase PCR (RT-PCR) analysis. For detailed expression analysis of GluR␦1 in the inner ear, laser capture microdissection was performed using the PixCell II system (Arcturus). We used a previously described method (32) with some modifications. Briefly, the mice were anesthetized and intracardially perfused with 4% paraformaldehyde in phosphate buffer. Temporal bones were removed, and oval windows were opened for the injection of fixative. Cochleas were then postfixed overnight and decalcified in 0.1 M EDTA for 1 to 3 days. The cochleas were dehydrated in ascending concentrations of alcohol and embedded in paraffin. The embedded cochleas were sectioned into 12-m thicknesses, and sections were mounted on uncharged slides (six sections on each slide). The sections were deparaffinized in xylene and dried at room temperature. We captured inner hair cells, outer hair cells, spiral ganglion cells, type I and IV fibrocytes, Deiters' cells, Claudius cells, Boettcher cells, inner sulcus cells, marginal cells, and vestibular hair cells from eight slides from each mouse. We pooled all of the cells in each category from different slides into a single tube. As a control, we scraped the whole sections from one slide into a tube. Three GluR␦1 ϩ/ϩ mice and one GluR␦1 Ϫ/Ϫ mouse were independently analyzed.
We used the Paradise whole-transcript RT reagent system (Arcturus, Mountain View, CA) to purify mRNA from both whole sections and laser-captured cells from cochlear sections. We made cDNA by reverse transcription from the mRNA using the same kit as above. For PCR, we designed four pairs of primers to amplify the cDNA of GluR␦1 (forward, 5Ј ACCTCCTGGAATGGGATGAT; reverse, 5Ј CCTCAGGCTTCTTGATGAGG), prestin (forward, 5Ј AGTGGCT GCCAGCATATAAA; reverse, 5Ј CGATGAGTACAGGCCAAACA), p27 (forward, 5Ј ATTGGGTCTCAGGCAAACTCT; reverse 5Ј GTTCTGTTGGC CCTTTTGTTT), and ␤-actin (forward, 5Ј AATTTCTGAATGGCCCAGGT; reverse, 5Ј TGTGCACTTTTATTGGTCTCAA). We used cDNA of P9 whole cochlea as a positive control and mRNA of P9 whole cochlea without reverse transcription as a negative control for PCR.
Assays of cochlear function. For auditory brain stem responses (ABR), distortion product otoacoustic emissions (DPOAE), and endolymphatic potential (EP) measurements, mice were anesthetized with xylazine (20 mg/kg intraperitoneally) and ketamine (100 mg/kg intraperitoneally). For ABR, needle electrodes were inserted at the vertex and pinna, with a ground near the tail. ABR potentials were evoked with 5-ms tone pips (0.5-ms rise-fall with a cos 2 onset envelope, delivered at a rate of 35/s). The response was amplified (10,000 times), filtered (100 Hz to 3 kHz), and averaged with an analog-to-digital board in a LabVIEW-driven data acquisition system. The sound level was raised in 5-dB steps from 10 dB below threshold to a 90-dB sound pressure level (SPL). At each sound level, 1,024 responses were averaged (with stimulus polarity alternated) using an artifact reject system, whereby response waveforms were discarded if the peak-to-peak amplitude exceeded 15 V. The threshold was defined by visual inspection of stacked waveforms as the lowest SPL at which coherent responses were detectable at a latency consistent across levels.
The DPOAE at distortion frequency 2f 1 -f 2 was recorded with a custom acoustic assembly consisting of two one-quarter-inch condenser microphones to generate primary tones of different frequencies (f 1 and f 2 , with f 2 /f 1 ϭ 1.2 and f 2 level 10 dB Ͻ f 1 level) and a Knowles miniature microphone (EK3103) to record ear canal sound pressure. Stimuli were generated digitally (National Instruments; catalog no. 6052E), and the maximum level of stimuli for DPOAE was 80 dB SPL. Ear canal sound pressure was amplified and digitally sampled at 4 s. Fast Fourier transforms were computed from averaged waveforms of ear canal sound pressure, and 2f 1 -f 2 DPOAE amplitude and the surrounding noise floor were extracted. Noise floors ranged from Ϫ25 to Ϫ5 dB SPL, depending on frequency. Isoresponse contours were interpolated from amplitude-versus-level functions performed in 5-dB steps of primary level.
For EP measurement (on a separate cohort of animals), the bulla was opened, exposing the cochlea, and the bone of the otic capsule over the basal turn was opened using a small knife. The spiral ligament and stria vascularis were left intact. A glass micropipette electrode (20 M⍀) filled with 150 mM KCl was introduced into the opening of the cochlea using a Kopf micropositioner. The signal was amplified 10-fold and read by custom software. The EP was considered valid if (i) there was a rapid rise in voltage (Ͼ25 mV per 3-m advance), (ii) the peak voltage remained stable for 30 s or more, and (iii) the voltage returned to 0 when the electrode was retracted.
Acoustic injury. For the evaluation of the vulnerability of GluR␦1 Ϫ/Ϫ mice to acoustic injury, mice were exposed, awake, and unrestrained, within cages suspended inside a small reverberant sound exposure box. The exposure stimulus was generated by a custom white-noise source, filtered (Brickwall filter with a 60-dB-octave slope), amplified (Crown power amp), and delivered (JBL compression driver) through an exponential horn fitted securely to a hole in the top of a reverberant box. Sound exposure levels were measured at four positions within each cage using a one-quarter-inch Bruel and Kjaer condenser microphone; sound pressure was found to vary by less than 0.5 dB across these measurement positions. Sound pressure was calibrated by positioning the microphone at the approximate position in relation to the animal's head. Mice were exposed for 2 h to the octave band noise (8 to 16 kHz) at 89 dB SPL.
Assays of vestibular function. For vestibular function in GluR␦1 Ϫ/Ϫ mice, we employed various behavioral and electrophysiological vestibular tests. In the Rotarod test, the rod (San Diego Instruments) rotated at speeds increasing in 5-rpm increments from 0 to 20 rpm and the retention time of the mice was recorded. Mice were tested in four trials each day at the same hours of the day over 4 days. For the swim test, each mouse was placed in a glass aquarium filled with tepid water. The time required for the mouse to surface and to maintain a horizontal bodyline swimming at the surface was recorded.
For linear vestibular evoked potential (VsEP) measurements, experimenters were blinded to the genotype during data collection and analysis. Animals were anesthetized (Equithesin, 4 l/g intraperitoneally), and each skull was prepared with a head mount. A thumbscrew was secured at the midline, and two additional electrodes were placed behind the left and right pinnae with a ground at the ventral neck. The animals were placed supine, and each head was secured to an electromechanical shaker with the naso-occipital axis oriented vertically. Stimuli were linear acceleration pulses (2-ms duration; 16 pulses/s) presented in two directions, normal and inverted. Normal polarity was defined as upward displacement, while inverted stimuli displaced the platform downward. Stimulus amplitude was measured in jerk (i.e., g/ms, where 1.0g ϭ 9.8 m/s 2 and 1.0g/ms ϭ 9.8 m/ms 3 [18]) using a calibrated accelerometer attached to the shaker platform. To monitor the jerk component of the stimulus, the output of the accelerometer was differentiated electronically. Stimulus amplitude was recorded in dB re: 1.0g/ms and ranged from Ϫ18 to ϩ6 dB re: 1.0g/ms, adjusted in 3-dB steps. Electrophysiologic activity was amplified (200,000-fold) and filtered (300 to 3,000 Hz), and VsEPs to normal and inverted stimulus polarities (1,024 points, 10 s/point, 128 responses per averaged waveform) were recorded. Four waveforms were obtained at each stimulus intensity level, two for normal polarity stimuli and two for the inverted polarity. Averaging of responses to normal and inverted polarities was completed offline to produce the final averaged waveforms used for analysis. Three response parameters were quantified: threshold, peak latencies, and peak-to-peak amplitudes. The threshold measured in dB re: 1.0 g/ms was defined as the stimulus amplitude midway between that which produced a discernible VsEP and that which failed to produce a response. Thresholds, latencies, and amplitudes were compared among the three groups of animals using one-way analysis of variance (ANOVA) (thresholds) and multivariate ANOVA (latencies and amplitudes) with a significance level of P less than 0.05.
Hippocampal electrophysiology. For the evaluation of the role of GluR␦1 in synaptic transmission and synaptic plasticity, hippocampal slices were prepared from GluR␦1 Ϫ/Ϫ and GluR␦1 ϩ/ϩ male mice without prior knowledge of mouse genotype. Slices were continuously superfused with artificial cerebrospinal fluid containing 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 2 mM MgSO 4 , 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , and 10 mM glucose, with 95% O 2 and 5% CO 2 at 30 to 31°C (2 ml/min). Schaffer collateral synapses were stimulated with a bipolar tungsten electrode in CA1 stratum radiatum placed 100 to 150 m from the recording pipette, and field excitatory postsynaptic potentials (fEPSPs) were collected using a MultiClamp 700B amplifier (Molecular Devices). To ensure equivalent activation of postsynaptic neurons in all experiments, stimulation intensities were chosen to evoke an fEPSP with a slope of approximately 1 mV/ms. In long-term potentiation (LTP) experiments, Schaffer collaterals were stimulated at 0.033 Hz before and after the induction of LTP. LTP was induced by a 200-Hz pulse protocol consisting of 10 trains of 200 ms of stimulation at 200 Hz delivered every 5 s at the baseline stimulation intensity. Data were analyzed using Clampfit 9.0 software (Molecular Devices). Results were grouped according to mouse genotype.

Morris water maze test. For an examination of defects in LTP in GluR␦1
Ϫ/Ϫ mice in vivo, mice of three genotypes were tested in a water maze that consisted of a circular blue plastic tank, 160 cm in diameter and 38 cm deep. The maze was located in a large test room surrounded by external cues that could be used for spatial navigation. The tank was filled to 30 cm with water at 21°C made opaque by the addition of a small quantity of nontoxic white paint (tempera). The platform, a 10-cm square of Plexiglas covered with a rough green plastic scouring pad, was mounted on a solid column 1 cm below the surface such that it could not be seen from water level. Four equally spaced points around the edge of the tank were used as start positions and divided the maze into four quadrants. During the acquisition of the place task, the platform was in the middle of one quadrant, equidistant between the center and the outer wall of the maze. Mice were trained for one block of four trials on each of 10 consecutive days. Within each block of trials, all four start positions were used once each in a pseudorandom sequence. For each trial, a mouse was placed in the water facing the wall at the start position. The time required to find the escape platform was recorded. Any mouse failing to find the platform within 60 s was placed on the platform. Approximately 10 min separated the individual trials in each day's block of tests.

RESULTS
Generation of GluR␦1 ؊/؊ mice. To create GluR␦1 Ϫ/Ϫ mice, we designed a targeting construct that deleted exons 11 and 12 of the GluR␦1 gene (Fig. 1A). This targeted disruption ensured the removal of three of the four transmembrane domains and introduced a frameshift after exon 12. We screened 380 ES cell colonies by genomic Southern blot analysis using an external probe, and one underwent homologous recombination. Using Neo as a probe, we confirmed that there were no other random integrations in this ES cell line (data not shown). We performed karyotyping to determine cytogenetic normality. After blastocyst injection, high chimeras were obtained, and germ line transmission was achieved. The crosses between GluR␦1 ϩ/Ϫ mice yielded offspring with an approximately 1:2:1 ratio of the GluR␦1 ϩ/ϩ (63 offspring), GluR␦1 ϩ/Ϫ (126 offspring), and GluR␦1 Ϫ/Ϫ (65 offspring) genotypes, suggesting no embryonic lethality in the GluR␦1 Ϫ/Ϫ mice. The correct targeting of GluR␦1 gene was further confirmed by genomic Southern blot analysis of mice with germ line transmission (Fig. 1B). In the PCR analysis, no 220-bp wild-type bands (in the deleted region) were detected in the homozygous mice (Fig. 1C).
To confirm the ablation of the GluR␦1 protein in GluR␦1 Ϫ/Ϫ mice, we performed Western blot analysis with a polyclonal antibody against the C termini of both the GluR␦1 and GluR␦2 proteins. As a positive control for this antibody, we used the cerebellum, where GluR␦2 is predominantly expressed and is unchanged in either GluR␦1 ϩ/ϩ or GluR␦1 Ϫ/Ϫ mice ( Fig. 2A); because GluR␦2 is not normally expressed in the inner ear or hippocampus (36), this antibody can be used to assay GluR␦1 expression in these tissues from mutant mice. Both GluR␦1 ϩ/ϩ and GluR␦1 ϩ/Ϫ mice expressed the expected ϳ115-kDa GluR␦1 protein, and no GluR␦1 protein with a mass of 115-kDa or any other size was detected in the inner ear or hippocampus of GluR␦1 Ϫ/Ϫ mice ( Fig.  2A). It remains possible that a small N-terminal peptide is made in GluR␦1 Ϫ/Ϫ ; however, it is likely that this peptide would be degraded due to the lack of proper transmembrane domains and a C terminus, so GluR␦1 Ϫ/Ϫ is therefore an effective null allele.
In addition, we performed immunofluorescence on hippocampus and cerebellum from GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice using a GluR␦1-specific antibody ( Fig. 2B; data not shown). GluR␦1 is absent in cerebellum but present in hippocampus of GluR␦1 ϩ/ϩ mice; however, it is absent in both regions of GluR␦1 Ϫ/Ϫ mice ( Fig. 2B; data not shown), consistent with Western blot results ( Fig. 2A). Furthermore, our wild-type immunostaining results are consistent with in situ hybridization results reported recently for developing and adult mouse brains (24,26). Our Western blotting and immunostaining results also demonstrated that GluR␦2 is absent in the inner ear and hippocampus and that no obvious up-regulation of GluR␦2 occurs in the cerebellum, hippocampus, or inner ear of GluR␦1 Ϫ/Ϫ mice. GluR␦1 Ϫ/Ϫ mice showed no obvious developmental or behavioral abnormality, except for overall weight reductions by 4 months of age. In all of the groups at 2 months of age or younger, weight differences were not significant (data not shown). At 4 months, the weights of male GluR␦1 Ϫ/Ϫ mice were approximately 89% of those of the GluR␦1 ϩ/ϩ males, and the weights of female GluR␦1 Ϫ/Ϫ mice were approximately 87% of those of the GluR␦1 ϩ/ϩ females; there was no significant difference between the weights of GluR␦1 ϩ/Ϫ and GluR␦1 ϩ/ϩ mice of either sex.
Cochlear GluR␦1 expression. To independently verify the expression of GluR␦1 in specific single cell types of the inner ear of GluR␦1 ϩ/ϩ mice and to confirm the deletion of GluR␦1 in the inner ear of GluR␦1 Ϫ/Ϫ mice, we used laser capture microdissection and RT-PCR (Fig. 2C). Prestin (an OHCspecific marker), p27 (a supporting cell marker), and ␤-actin (a ubiquitous marker) were used as controls. GluR␦1 was expressed in IHCs, OHCs, spiral ganglia, and vestibular HCs. (Fig. 2C). Our results are largely consistent with those of a previous report (36); however, some differences exist: in previous studies, OHCs were negative in both in situ and immunostaining analyses for both guinea pigs and rats, whereas Claudius cells in basal turns were negative in rat cochleae by in situ analysis but weakly positive in guinea pig cochleae by immunostaining analysis with only anti-GluR␦1 antibody (see Fig. 2 and 5 in reference 36). Such differences are subtle and can be subject to differences between species or in sensitivities of detection methods. In the inner ears of GluR␦1 Ϫ/Ϫ mice, the corresponding portion of GluR␦1 mRNA was indeed deleted in all cell types analyzed (Fig. 2C). These results were reproduced in three independent experiments using different GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mouse cochlear sections.

However, it was not expressed in type I and IV fibrocytes, Deiters' cells, Claudius cells, inner sulcus cells, Boettcher cells, or marginal cells in stria vascularis
Cochlear responses. Given the strong cochlear expression of GluR␦1, we examined cochlear function in GluR␦1 Ϫ/Ϫ mice. The three assays used were (i) ABR, the summed soundevoked activity of the auditory nerve and ascending auditory neural pathways; (ii) DPOAE, a sound-evoked preneural signal generated and amplified by the OHCs and transmitted back to the ear canal; and (iii) the magnitude of the EP, the potential generated by the stria vascularis and measured inside the endolymphatic space, which generates the transepithelial electric driving force necessary to drive transduction currents through the hair cell stereocilia when they are deflected by acoustic stimulation.
EP measured in the basal turn (Fig. 3C), a cochlear location corresponding to the 45-kHz tonotopic location (17), was lower in GluR␦1 Ϫ/Ϫ mice (77.1 Ϯ 8.2 mV [mean Ϯ standard error of the mean {SEM}]; n ϭ 7) than in GluR␦1 ϩ/ϩ (100.7 Ϯ 3.2 mV [mean Ϯ SEM]; n ϭ 7) and GluR␦1 ϩ/Ϫ mice (90.0 Ϯ 6.2 mV [mean Ϯ SEM]; n ϭ 7). The difference between GluR␦1 Ϫ/Ϫ and GluR␦1 ϩ/ϩ mice was significant (P was Ͻ0.01 by Student's t test; P was Ͻ0.05 by one-way ANOVA). show ABR and DPOAE data, respectively, for the same cohort of animals; panel C is from a separate group. Means and standard errors (error bars) are shown; numbers of animals ("n") in each group are given. Statistical analyses are described in the text. Arrows on DPOAE points from GluR␦1 Ϫ/Ϫ mice indicate that these values are minimum estimates because some ears showed no response at the highest sound levels (80 dB SPL). The double asterisks in panel C represent a significant difference (P was Ͻ0.01 by Student's t test; P was Ͻ0.05 by one-way ANOVA) between GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice, while other group comparisons showed insignificant differences.
Cochlear morphology and immunostaining. To evaluate morphological changes in the GluR␦1 Ϫ/Ϫ mice, we examined plastic sections of osmium-stained cochleas. Histologic staining in GluR␦1 Ϫ/Ϫ mice (8 weeks old) showed a pathological pattern consisting of variable and scattered OHC loss in the basalmost region of the cochlea (Fig. 4E) and consistent and substantial loss of type IV fibrocytes in the spiral ligament (Fig.  4D) throughout much of the basal turn (Fig. 4E). As shown in high-power micrographs (Fig. 4B and D), the nuclei of type IV fibrocytes are normally visible interspersed among a complex fibrous network visible in differential interference contrast optics (Fig. 4D). In the mutant, all cell nuclei are absent in this region of the spiral ligament, and only the fibrous network remains (Fig. 4B). According to a mouse cochlear frequency map (8), the OHC loss was restricted to cochlear frequency regions well above that where the threshold shift was seen; however, the region of spiral ligament pathology correlated well with the region of threshold shift. Quantitative results from one ear of each genotype are shown in Fig. 4E to F: similar results for the loss of fibrocytes were seen in the other ears evaluated (n ϭ 3 of each genotype). There was no loss of IHCs or cochlear neurons, and all other structures of the cochlear duct, including the stria vascularis, appeared normal.
To further evaluate the condition of remaining cells of the spiral ligament and stria vascularis, we performed immunostaining for markers normally expressed in these areas and implicated in ionic homeostasis and therefore EP generation: Kv3.1b, Kir4.1, Atp1b1, and Nkcc1 (7,10,14,38). There was little evidence of down-regulation of these key channels/pumps, except that Kv3.1b staining was reduced or absent in regions where type IV fibrocytes were missing in the basal turns but remained present in the type IV fibrocytes in the apical and middle turns and even in adjacent cells in the basal turns ( Fig. 5A and B; data not shown). Similarly, Kir4.1 and Atp1b1 appeared normal in marginal cells of the stria vascularis (Fig. 5C, D, G, H). Kir4.1 expression also appeared normal in the Deiters' cells, the supporting cells for OHCs, in both GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice (data not shown). No significant change in Nkcc1 staining was observed in marginal cells, type IV fibrocytes, and other cochlear cells of GluR␦1 Ϫ/Ϫ or GluR␦1 ϩ/ϩ mice (Fig. 5E and F; data not shown). Cochlear vulnerability to noise damage. Synaptic transmission between the IHC and its afferent innervation is glutamatergic, and acoustic overstimulation produces a type of glutamate excitotoxicity that can contribute to temporary noise-induced threshold shifts after acoustic overstimulation. In search of a functional role for the GluR␦1 receptor in the IHC area, we examined the vulnerability of GluR␦1 Ϫ/Ϫ mice to temporary acoustic injury. Twelve hours after a 2-h exposure to a noise band (8 to 16 kHz) at 89 dB SPL, GluR␦1 ϩ/ϩ mice displayed threshold shifts in both ABR (Fig. 6A) and DPOAE (Fig. 6B) responses that recovered within 1 week (data not shown). In GluR␦1 Ϫ/Ϫ mice, the temporary threshold shift was much larger for the same exposure, and the increased dysfunction was seen in both ABR and DPOAE (Fig. 6A and B). The symmetry of the threshold shifts in both the neural (ABR) and the preneural, OHC-based measures (DPOAE) suggests that the shifts are well explained by OHC dysfunction and, thus, that the increased vulnerability did not arise exclusively in the IHC area, i.e., it is not due to enhanced excitotoxicity at the IHC/afferent synapse (29). In contrast, when the efferent innervation of the IHCs is selectively destroyed, the increased vulnerability is seen only in the ABRs and not the DPOAEs, consistent with changes in synaptic transmission associated with excitotoxicity (5). Cochlear efferent innervation and efferent inhibition. The vulnerability of the cochlea to acoustic injury is controlled, in part, by a cholinergic feedback inhibitory circuit to the OHCs, the medial olivocochlear pathway (27). Given the enhanced vulnerability of the GluR␦1 Ϫ/Ϫ mice to acoustic injury, we evaluated the integrity of the efferent innervation, both morphologically and functionally.
To assess the density of efferent innervation, we immunostained cochlear whole mounts for SNAP25 (Fig. 7A) or synaptophysin (synaptic vesicle-associated proteins abundant in efferent terminals) or CGRP or VAT (markers for neurotrans-mitters found in efferent terminals; data not shown). Qualitative analysis suggested no abnormalities in the efferent innervation of the GluR␦1 Ϫ/Ϫ ears. In one cochlea from each genotype, we quantified SNAP25 immunostaining in both IHC and OHC areas and found no significant differences ( Fig. 7B and C).
To assess the function of the OHC efferent pathway, we measured the suppression of the DPOAEs elicited by electrical stimulation of the efferent bundle at the floor of the IVth ventricle (Fig. 7D). There were no significant differences among the three genotypes in the mean magnitude of this FIG. 7. There is no change in the density of efferent innervation (A to C) or in the strength of shock-evoked efferent effects (D and E) in GluR␦1 Ϫ/Ϫ mice. Panel A shows SNAP25 immunostained efferent terminals in the ear of a GluR␦1 ϩ/ϩ mouse. The white arrow marks a terminal in the OHC area, the black-rimmed arrow shows a terminal in the IHC area. Panels B and C show the total silhouette area of immunostained terminals as a function of cochlear location (converted to frequency) in one ear from each genotype. In the OHC area (C), values are averaged over all three OHC rows. Panel D shows representative data for each genotype from the assay used to measure cochlear efferent suppression. DPOAE amplitude (evoked with f 2 at 16 kHz) was measured repeatedly before, during, and after a train of shocks to the efferent bundle; efferent suppression of DPOAE amplitude (the difference between the two dashed lines) was measured. Panel E shows average values of efferent suppression at different test frequencies obtained from a cohort of animals using the assay illustrated in panel D. Means and standard errors (error bars) are shown. ϩ/ϩ, GluR␦1 ϩ/ϩ ; Ϫ/Ϫ, GluR␦1 Ϫ/Ϫ ; ϩ/Ϫ, GluR␦1 ϩ/Ϫ . VOL. 27,2007 GluR␦1 REQUIRED FOR HIGH-FREQUENCY HEARING 4507 efferent effect (Fig. 7E), except at 32 kHz, where the reduced efferent effect in the GluR␦1 Ϫ/Ϫ mice is well explained by the OHC dysfunction in that region, as seen by the threshold elevation in both ABRs and DPOAEs ( Fig. 3A and B). Since the efferent suppressive effect on cochlear sound-induced vibration arises by reducing the OHCs' contribution to cochlear amplification, efferent effect size is always reduced in areas of OHC dysfunction. We conclude that the deletion of GluR␦1 did not affect efferent synaptic transmission and that efferent dysfunction cannot account for the enhanced vulnerability to acoustic injury. Vestibular morphology and function. The presence of GluR␦1 in vestibular hair cells (both type I and type II) and vestibular ganglia (36) suggested a role in vestibular function. We first analyzed morphology of vestibular end organs in osmicated plastic sections and hematoxylin-and-eosinstained paraffin sections (see Materials and Methods). There were no changes in morphology of the saccule, utricle, or semicircular canals or in their afferent innervation in the GluR␦1 Ϫ/Ϫ mice (Fig. 8A to D; data not shown). Ves-tibular function was measured at 2 months, both behaviorally (Rotarod and swim tests) and electrophysiologically (VsEP, the summed neural activity of vestibular afferents from the utricle and saccule, and the ascending vestibular pathway, evoked by linear acceleration stimuli). The loss of GluR␦1 had no significant effect on the ability of mice to remain on a rotating rod as its speed of revolution increased (Fig. 8E) or on the time required to right themselves and begin swimming after being dropped into a water bath (data not shown). On average, VsEP thresholds were slightly higher in GluR␦1 Ϫ/Ϫ mice than in GluR␦1 ϩ/ϩ mice. However, differences among the groups were not statistically significant (ANOVA) (Fig. 8F). Similarly, P1 peak latency, P2 peak latency, P1-N1 amplitudes, and P2-N2 amplitudes were not significantly different among the three genotypes (multivariate ANOVA). All response parameters were similar to normative values in normal young and adult mice (18,19). These data suggest that the absence of the GluR␦1 does not significantly alter gravity receptor function or balance behavior. in the Rotarod test did not reveal significant differences among mice of three genotypes at 2 months of age. The number of mice tested in each group is illustrated. (F) Threshold measurements (mean and SEM [error bars]) of linear VsEPs did not show significant differences among mice of three genotypes at ages of between 2 and 3 months. The stimulus amplitude was described in dB re: 1.0g/ms and ranged from Ϫ18 to ϩ 6 dB re: 1.0g/ms, adjusted in 3-dB steps. The number of mice tested in each group is illustrated. ϩ/ϩ, GluR␦1 ϩ/ϩ ; Ϫ/Ϫ, GluR␦1 Ϫ/Ϫ ; ϩ/Ϫ, GluR␦1 ϩ/Ϫ .
Hippocampal morphology and function. Because of the high level of GluR␦1 mRNA and protein in adult hippocampus (Fig. 2B) (25,36), we compared hippocampal morphology and synaptic function in GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice at 2 months of age. As shown in Fig. 9A, no gross morphological differences were detected among the genotypes. Because the deletion of GluR␦2 strongly affects longterm depression of synaptic transmission in the cerebellum, we examined the role of GluR␦1 in synaptic transmission and synaptic plasticity at excitatory synapses between CA3 and CA1 pyramidal neurons (CA3-CA1 synapses) in hippocampal slices ( Fig. 9B and C). Recordings of the extracellular fEPSP showed that the loss of GluR␦1 did not cause any significant changes in synaptic transmission (Fig. 9B). Thus, input-output curves recorded over a wide range of stimulus intensities were normal in GluR␦1 Ϫ/Ϫ mice compared to their GluR␦1 ϩ/ϩ littermates. We next explored the effect of GluR␦1 deficiency on LTP at CA3-CA1 synapses.
We chose a 200-Hz stimulation protocol that induced a compound LTP that consisted of both presynaptic and postsynaptic modules of LTP expression (47). We found no significant changes in compound LTP between GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice (Fig. 9C). Thus, the 200-Hz tetanic stimulation fEPSPs was increased to 159 Ϯ 13% of initial levels in slices from GluR␦1 Ϫ/Ϫ mice (mean Ϯ SEM; n ϭ 10) and 174 Ϯ 17% in GluR␦1 ϩ/ϩ mice (n ϭ 15; P was Ͼ0.05 by the Kolmogorov-Smirnov test) 60 min after tetanization. Similarly, fEPSPs measured at 10 or 90 min after 200-Hz tetanization were not significantly different in GluR␦1 Ϫ/Ϫ mice compared to those in their GluR␦1 ϩ/ϩ littermates (P was Ͼ0.05 by the Kolmogorov-Smirnov test). The Morris water maze was used to test hippocampal function in vivo. As shown in Fig. 9D, mice of all genotypes learned the task as indicated by the steady decline in latency to find the hidden platform (ANOVA; day; F ϭ 20.25, df ϭ 9,216, P Ͻ 0.001). There were no significant group differences in the FIG. 9. Hippocampal morphology and function appear normal in GluR␦1 Ϫ/Ϫ mice. (A) Hematoxylin and eosin staining of hippocampi in GluR␦1 ϩ/ϩ (ϩ/ϩ) and GluR␦1 Ϫ/Ϫ (Ϫ/Ϫ) mice at 2 months old. No obvious differences in neuronal location, number, and position were detected between the hippocampuses of GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice. Synaptic transmission appeared normal in hippocampal slices of GluR␦1 ϩ/ϩ and GluR␦1 Ϫ/Ϫ mice at 2 months old (B and C). (B) Recordings of the extracellular fEPSPs showed that the loss of GluR␦1 did not cause any significant changes in synaptic transmission over a wide range of stimulus intensities. (C) A 200-Hz tetanic stimulation enhanced the fEPSPs to 159 Ϯ 13% (mean Ϯ SEM [error bars]; n ϭ 10) and 174 Ϯ 17% (n ϭ 15; P was Ͼ0.05 by the Kolmogorov-Smirnov test) of their initial levels, when measured 60 min after tetanization in slices from GluR␦1 Ϫ/Ϫ and GluR␦1 ϩ/ϩ mice, respectively. (D) Performance in the place task of the water maze did not show significant differences among GluR␦1 ϩ/ϩ , GluR␦1 ϩ/Ϫ (ϩ/Ϫ), and GluR␦1 Ϫ/Ϫ mice at 2 to 3 months of age (n ϭ 6, 10, 11, respectively). Mean latency to reach the hidden platform is shown on each of the 10 test days. All groups performed similarly in this task. The vertical bar indicates one standard error of the difference (SED) in group means.

DISCUSSION
Targeted disruption of GluR␦1 causes significant hearing loss at high frequencies, associated with reductions of both OHC function and EP, the resting potential of the lumen of the cochlear duct that helps drive receptor currents into sensory cells. These findings provide the first in vivo evidence of a functional role for this largely uncharacterized orphan glutamate receptor, which is consistent with its prominent inner ear expression. Given the prevalence of congenital or acquired high-frequency hearing loss in human ears, the locus encoding GluR␦1 represents a candidate disease gene (39).
In contrast, the apparently normal function and morphology in vestibular sensory-end organs and hippocampus in GluR␦1 Ϫ/Ϫ mice suggest that functional redundancy exists for the GluR␦1 expressed in these areas. Given the lack of other members of the GluR␦ family in human and mouse genome sequences, it is conceivable that other proteins with little sequence homology to GluR␦1 may compensate for the lack of GluR␦1 in vestibular end organs and hippocampus.
Cochlear dysfunction and EP reduction. The high-frequency threshold elevation observed in the GluR␦1 Ϫ/Ϫ mice was on the order of 20 to 45 dB as measured in the neural response (ABR) and roughly half that as measured in the preneural response (DPOAE). This hearing loss was associated with a reduction in the EP of 20 to 25 mV, as measured in the basal turn at roughly the 45-kHz place. A number of lines of evidence converge to suggest that both the pattern and degree of threshold elevations are well explained qualitatively and quantitatively by the EP reduction (29), i.e., there is no reason to assume dysfunction in the sensory cells and nerve fibers per se.
Decreasing the EP reduces the driving force for soundelicited transduction currents into both IHCs and OHCs. This reduction seen by the OHCs reduces their somatic electromotility and thereby decreases mechanical motions of the cochlear partition and elevates thresholds as seen by both DPOAEs and ABRs (35). Threshold elevation in cochlear neurons is further increased by effects at the IHCs, since these IHCs provide exclusive synaptic drive to 95% of these nerve fibers. The EP reduction seen by IHCs reduces the receptor potentials which drive synaptic transmission, even without a change in cochlear vibration. Thus, cochlear dysfunction from EP reduction results in larger changes in ABR (neural) thresholds than in DPOAE thresholds (which require only normal OHC function). Indeed a recent empirical comparison of ABR shifts and DPOAE shifts in furosemide-treated gerbils showed a ratio very similar to that seen here (29).
According to studies of click-evoked neural potentials in cats, there is roughly a 1-dB threshold shift in ABR for each 1-mV decrement in EP (37). Given that click-evoked thresholds in cat are dominated by midfrequency (5 to 10 kHz) neurons (1) and that the OHC contribution to the cochlear amplifier increases with frequency, it is not unlikely that, in the 16-to 45-kHz region of the mouse, the relationship between neural thresholds and EP reduction is greater than 1 dB/mV. Thus, the ABR and DPOAE shifts seen in the present study are well explained by the magnitude of the EP shift.

EP reduction and loss of spiral ligament cells.
The EP is generated by coordinated ion pumping activity of numerous cell types in the spiral ligament and the stria vascularis. Numerous other deafness mutations appear to affect hearing via their effects on EP and cochlear ion homeostasis. Mutations in Cx26, Claudin-11, Pendrin, Claudin-14, Nkcc1, Kcc4, and various channels (e.g., Kir4.1 and Isk) can cause EP reduction and corresponding elevation of cochlear thresholds (2-4, 7, 11, 12, 22, 28, 41). Mice lacking Pou3f4, a transcription factor expressed in spiral ligament fibrocytes, showed a 50-mV EP reduction associated with a 70-to 80-dB elevation of ABR thresholds. Interestingly, such a profound cochlear dysfunction (much more dramatic than that seen here with the loss of GluR␦1) was associated with very subtle morphological changes. Hair cells and the rest of the organ of Corti were normal in Pou3f4 mice, with ultrastructural changes noted only in spiral ligament fibrocytes (type I and type II) (30). Mice lacking otospiralin, a protein of unknown function produced by spiral ligament fibrocytes, displayed modest threshold elevation (20 dB by ABR) associated with subtle changes in morphology of type II and type IV fibrocytes, visible only at the ultrastructural level. EP was not measured in this mutant line (6).
Type IV fibrocytes in GluR␦1 Ϫ/Ϫ mice were eliminated throughout the basal turn. However, this fibrocyte loss was probably not the cause of the EP reduction. Acoustic overstimulation experiments in mice have shown that type IV fibrocytes, among the most vulnerable cells in the ear, can be eliminated after moderate noise exposures, yet ABR thresholds and EP values can completely recover (17,42). Similarly, in a mouse (C57BL/6) with progressive, high-frequency agerelated hearing loss, type IV fibrocytes are also among the first cells to disappear from the basal turn (13) and, although highfrequency thresholds are elevated, the EP is not reduced (23, 31) (K. Hirose and M. C. Liberman, unpublished data).
Given that GluR␦1 was expressed in IHCs, OHCs, and spiral ganglion neurons but not in the spiral ligament, possible explanations for the EP shifts based on cellular changes outside the stria and ligament must be considered. One possible link is that the generation of a normal EP must depend on appropriate recycling of K ϩ from the hair cells to the stria via the spiral ligament (43), and the loss of GluR␦1 may disrupt that recycling in either the IHC or the OHC areas. The organ of Corti is both mechanically labile and "leaky" to ion flux, as a result of the effects of the loss of GluR␦1 on one or more supporting cells. Direct measurement of the input impedance of scala media may address this issue, although such an approach is tedious and artifact prone.
Glutamatergic transmission and loss of OHC function. Although there is a rich efferent innervation of hair cells and cochlear neurons, there is no evidence for glutamatergic synapses in the efferent system; correspondingly, our findings showed no effects of GluR␦1 deletion on efferent innervation or the strength of efferent-evoked effects on cochlear response (i.e., DPOAEs). However, the striking increase in vulnerability of the ears of GluR␦1 Ϫ/Ϫ mice to temporary acoustic injury seen in this study was consistent with a role of GluR␦1 in OHCs. In particular, the similarity in noise-induced DPOAE shifts and ABR shifts suggests that increased vulnerability is occurring presynaptically, e.g., involving OHCs and their role as cochlear amplifiers. The presence of GluR␦1 in OHCs in 4510 GAO ET AL. MOL. CELL. BIOL.
our study and others (36) is consistent with such a notion. The afferent synapse between OHCs and type II cochlear nerve fibers is poorly understood. It is not clear whether afferent transmission there is glutamatergic (33,34); glutamate excitotoxicity is not seen in the OHC area after acoustic overstimulation (33,34), and the efferent synapses on OHCs are clearly cholinergic in nature with end effects mediated via the ␣9/␣10 nicotinic acetylcholine receptors (9,40). Thus, it is difficult to propose a compelling argument as to why GluR␦1 loss should enhance damage to OHCs per se. An alternate hypothesis is that the heightened vulnerability arises via increased fragility of the EP generation mechanisms such that the increased demands on the system imposed during acoustic overstimulation lead to further EP reductions not seen in mice with more robust ion homeostasis.