Tuning of the Outer Hair Cell Motor by Membrane Cholesterol*

Cholesterol affects diverse biological processes, in many cases by modulating the function of integral membrane proteins. We observed that alterations of cochlear cholesterol modulate hearing in mice. Mammalian hearing is powered by outer hair cell (OHC) electromotility, a membrane-based motor mechanism that resides in the OHC lateral wall. We show that membrane cholesterol decreases during maturation of OHCs. To study the effects of cholesterol on hearing at the molecular level, we altered cholesterol levels in the OHC wall, which contains the membrane protein prestin. We show a dynamic and reversible relationship between membrane cholesterol levels and voltage dependence of prestin-associated charge movement in both OHCs and prestin-transfected HEK 293 cells. Cholesterol levels also modulate the distribution of prestin within plasma membrane microdomains and affect prestin self-association in HEK 293 cells. These findings indicate that alterations in membrane cholesterol affect prestin function and functionally tune the outer hair cell.

Cholesterol is an important component of the plasma membranes of most animal cells. It modulates the mechanical properties of the membrane and affects the function of membraneassociated proteins. Recent studies have shown modulation by membrane cholesterol of such diverse membrane proteins as rhodopsin (1,2), the serotonin receptor 1A (3) and serotonin transporter 5HT1 (4), the chloride channel ClC-2 (5), several classes of potassium channels (6,7), the nicotinic acetylcholine receptor (8), and several G-protein-coupled receptors (9,10).
Cellular cholesterol levels are tightly regulated, and disruption of cholesterol homeostasis leads to a host of disease conditions. Several clinical and experimental studies carried out using rabbits, chinchillas, guinea pigs, and human subjects have linked sensorineural hearing loss and/or increase of hearing thresholds to hypercholesterolemia (27)(28)(29)(30)(31). Reduction of cholesterol by statins or apheresis has been shown to delay hearing loss in mice (32) and improve hearing recovery in humans (33). A study of hypercholesterolemic humans indicated that the effects of cholesterol on hearing might involve effect(s) on nonlinear mechanical processes in the cochlea (34).
A piezoelectric-like membrane-based motor in the outer hair cell (OHC) 7 contributes to the exquisite sensitivity and frequency selectivity of mammalian hearing. This motor mechanism is required to counteract viscous damping in the fluidfilled cochlea, which would otherwise impair mechanical tuning. The OHC lateral wall is specialized for electro-mechanical force transduction (35,36). Here, the energy in the transmembrane electric field is converted into mechanical energy. The organization of the OHC lateral wall is unique among hair cells and among all adult mammalian cell types. It is an elegant, nanoscale (ϳ100 nm thick), trilaminate structure. The outer and inner layers are the plasma membrane and subsurface cisternae, respectively, and sandwiched between them is a layer of cytoskeletal proteins called the cortical lattice. The lipid composition of the plasma membrane and the subsurface cisternae membranes is unknown, but the constituent lipids are in the fluid phase allowing for free diffusion (37)(38)(39). Labeling studies suggest that lateral wall membranes contain less cholesterol than the OHC apical and basal plasma membranes (40 -43). The relatively low cholesterol level of the OHC lateral wall plasma membrane is unusual among animal cells, and may serve to modulate the function of the membrane proteins that reside there. These proteins include a modified anion exchanger AE2 (44), the Glut5 sugar transporter (45)(46)(47), stretch-activated ion channels (48 -50), and prestin (45).
Prestin (SLC26A5), a critical component of the OHC lateral wall motor, is a polytopic integral membrane protein (45,51,52) and is essential for OHC electromotility and mammalian hearing (53). Prestin greatly increases charge movement into and out of, as opposed to through, the membrane (54,55). Intracellular anions such as chloride and bicarbonate have been shown to be the charge carrier (55) consistent with the membership of prestin in the SLC26A family of anion transporters (56). When transfected into several mammalian cell lines, prestin confers a voltage-dependent nonlinear capacitance (NLC), the accepted electrical signature of electromotility (54, 55) (see supplemental text for in-depth description).
Motivated by the clinical effects of cholesterol on hearing and the reduced cholesterol levels in the OHC lateral wall, we have explored the effect of cholesterol on hearing at the organ, cellular, and molecular levels to clarify its biological basis of action. We observe that cholesterol affects otoacoustic emissions and functionally tunes nonlinear mechanical processes in the OHC, most likely through its effects on the OHC membrane protein prestin.

DPOAE Measurements
Mice used for DPOAE measurements were of a mixed genetic background derived from two strains, 129SvEv and C57B6/J, and were 4 -8 weeks old. Healthy mice were anesthetized with ketamine/xylazine and immobilized in a head holder. The pinna was resected and the middle ear bulla opened to expose the round window. An earbar connected to two speakers and a probe tip microphone were inserted into the ear canal to within 2 mm of the tympanic membrane. The cubic distortion product amplitude was measured using an F2 frequency of 20 kHz with F1 ϭ F2/1.2 (57). The intensities of the primary tones were equal. First, we ranged the primary tones from 20 to 80 db in 10-db steps to verify that there was no notch in the DPOAE amplitude curve between 50 and 70 db. During the experiment, we set the primary tones to 60-db sound pressure level, and the DPOAE amplitude was measured every 9 s. After a few minutes, a borosilicate micropipette with a tip diameter of ϳ50 m containing the treatment solution (either 100 mM M␤CD, 200 mM water-soluble cholesterol, ϳ200 mM raffinose, or 10 mM water-soluble cholesterol) was carefully inserted through the round window membrane. The high concentrations of each treatment (in comparison with established in vitro studies) were chosen to compensate for dilution of the solutions in the mouse perilymph. The treatment solutions were allowed to diffuse passively into the perilymph. The middle ear space was monitored for fluid seepage, and any fluid was carefully aspirated. DPOAE amplitudes were collected for up to 30 min. In some cases, at the conclusion of the experiment, the basilar membrane was perforated to eliminate DPOAEs, thereby verifying the measurements obtained. DPOAE amplitudes were then normalized so that the amplitude after the micropipette was inserted and all middle ear fluid was cleared was 0 db. This time window is indicated as a gray box in each panel of Fig. 1.

Outer Hair Cell Isolation
Albino guinea pigs of either sex weighing 200 -300 g and having a normal startle response to a hand clap were decapitated. The temporal bones were taken and the middle ear bullae opened. The otic capsule was removed, and the spiral ligament was peeled off to expose the organ of Corti. The modiolus with the intact organ of Corti was removed from the temporal bone and subjected to mild trypsinization for ϳ10 min at room temperature and trituration to detach OHCs. OHCs were plated onto the glass bottom of a coated microwell Petri dish (MatTek, Ashland, MA). Isolated cells were selected for study on the basis of standard morphological criteria within 4 h of animal death. Under the light microscope, healthy cells display a characteristic birefringence, a uniformly cylindrical shape without regional swelling, a basally located nucleus, and no Brownian motion of subcellular cytoplasmic particles (58).

Prestin Constructs and Transfection
Gerbil prestin was cloned into the pIRES-hrGFP vector (Stratagene, La Jolla, CA) as a HA tag fusion protein (HA-prestin) and into the pEGFP, pECFP, and pEYFP vectors (Clontech) as a GFP, CFP, or YFP fusion protein (prestin-EGFP), as described previously (59 -61). The prestin-ECFP and prestin-EYFP constructs were modified by site-directed mutagenesis (QuikChange mutagenesis kit, Stratagene, La Jolla, CA) to include a single amino acid substitution (A206K) on the CFP/ YFP fusion protein, which renders CFP/YFP monomeric. The sequences of the constructs were verified using five overlapping sequencing primers. NLC measurements confirmed that all constructs used in this study are functional in HEK 293 cells. HEK 293 cell lines were transfected 24 h after passage with prestin-EGFP, prestin-ECFP, prestin-EYFP, or HA-prestin at a 3:1 ratio of DNA with FuGENE 6 (Roche Applied Science).

Cholesterol Manipulations
Outer Hair Cells-Because of the sensitivity of outer hair cells to temperature and their deterioration with time after isolation, cholesterol manipulations were performed differently in OHCs than in HEK 293 cells. Cholesterol depletion was carried out by pipetting M␤CD into the external solution in the dish containing hair cells at a final concentration of ϳ100 M (1/100th that used in HEK cells, see below) and incubating at room temperature (see Fig. 3 for times of incubation). Higher concentrations of M␤CD produced drastic morphological changes in OHCs and even cell death because of destabilization of the cholesterol-rich apical and basal membranes, causing the nucleus to be blown out of the cell. Cholesterol was loaded in a similar manner at a final concentration of 1 mM of M␤CD containing cholesterol (also referred to as water-soluble cholesterol). In both cases, treatment was carried out after forming a whole-cell patch on an OHC, and capacitance recordings were taken throughout the incubation time.

Electrophysiological Measurements
Electrophysiological data were obtained from cells using the whole-cell voltage clamp technique. Our recording techniques are fully described earlier (60) and a brief description follows. Culture dishes containing transfected cells were placed on the stage of an inverted microscope (Carl Zeiss, Gottingen, Germany) under ϫ100 magnification and extensively perfused with the extracellular solution containing Ca 2ϩ and K ϩ channel blockers prior to recording. All recordings were conducted at room temperature (23 Ϯ 1°C). Patch pipettes (quartz glass) with resistances ranging from 2 to 4 megohms were fabricated using a laser-based micropipette puller (P-2000, Sutter Instrument Company, Novato, CA) and filled with an intracellular solution, also containing channel blockers. For cell membrane admittance, Y was measured with the patch-clamp technique in the whole-cell mode using a DC voltage ramp with dual frequency stimulus (62) from Ϫ0.14 to 0.14 V with a holding potential of 0 V, and the cell parameters were calculated from the admittance as described earlier (63). The conductance, b, was also determined experimentally with a DC protocol, as described earlier (60).
In all representations, capacitances were normalized with respect to base-line capacitance (taken as the capacitance at 0.1 V), and peak capacitance (differs according to treatment), as in Equation 1, where C(V) is the capacitance at voltage V; C baseline is capacitance at base-line voltage (defined above), and C normpkc is equal to C norm at V pkc .

Tissue Preparation and Filipin Labeling of Mouse OHCs
P6, P12, and adult ICR mice were sacrificed by cervical dislocation and decapitation. The temporal bone was removed, and the bony capsule was stripped in fresh cold Hanks' balanced salt solution (Invitrogen). The membranous labyrinth was exposed in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. The sensory epithelium was isolated and affixed to round glass coverslips coated with Cell-Tak TM (BD Biosciences). The tissue was washed twice with PBS, fixed with 4% paraformaldehyde for 30 min, and stained with filipin dye (4 mg/ml) and AlexaFluor 594 phalloidin for 30 min. The samples were then washed twice with PBS, mounted on glass slides with Fluoromount G antifade reagent, and sealed with nail polish. Images were captured on a Zeiss Axioplan microscope (Carl Zeiss Optics Company, Jena, Germany) with ϫ63 objective and analyzed with Applied Precision SoftWoRx deconvolution software. Images of individual OHCs were analyzed using NIH Image software, and pixel intensities along a line drawn through the middle of a single OHC were plotted as bar graphs in Fig. 2.

Immunofluorescence and Imaging
HA-prestin transfected cells on coverslips were either treated with or without 10 mM M␤CD or water-soluble cholesterol for 30 min at 37°C. Cells were then washed with PBS, stained with concanavalinA-AlexaFluor 350 conjugate (Molecular Probes, Carlsbad, CA) for 1 h on ice, washed with PBS again, and then permeabilized with PBS/Triton X-100 before fixing with 4% paraformaldehyde in PBS. The cells were then stained with anti-HA antibody (1:1200; Cell Signaling Technology, Inc., Danvers, MA), followed by AlexaFluor 594 goat antimouse secondary antibody (1:800; Molecular Probes, Carlsbad, CA). Coverslips were mounted inverted on glass slides with Fluoromount G antifade reagent (Electron Microscopy Sciences, Hatfield, PA) and fluorescent images captured on a Zeiss LSM 510 deconvolution microscope (Carl Zeiss Optics Company, Jena, Germany) with ϫ63 objective and analyzed with Applied Precision SoftWoRx image restoration software. Images were also obtained using a Zeiss LSM 510 confocal microscope with ϫ63 objective and analyzed using Zeiss AIM imaging software.

Membrane Fractionation
Cell membranes were fractionated as described by Vetrivel et al. (64). Briefly, HEK 293 cells expressing HA-prestin, treated with or without M␤CD or with water-soluble cholesterol (as detailed above), were lysed in buffer (0.5% Lubrol WX, 25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, and 1 mM phe- DECEMBER 14, 2007 • VOLUME 282 • NUMBER 50 JOURNAL OF BIOLOGICAL CHEMISTRY 36661 nylmethanesulfonyl fluoride). Membranes were fractionated on a 5, 35, and 45% sucrose step gradient. Twelve 1-ml fractions were collected, and excess lipids in each fraction were removed by methanol/chloroform precipitation before the proteins were analyzed by 7.5% SDS-PAGE. This experiment was repeated at least six times with invariant results; a representative blot is shown in Fig. 6.

Cross-linking
Forty eight hours post-transfection with HA-prestin, HEK 293 cells were either treated with methyl-␤-cyclodextrin (M␤CD) or with water-soluble cholesterol for 30 min at 37°C, or left untreated, before gentle harvesting by scraping into 1 ml of PBS, pH 8.0. The cells were pelleted (2000 ϫ g for 5 min) and incubated with various concentrations of cross-linker bis(sulfosuccinimidyl) suberate (0.078 to 5 mM of BS 3 ) or without cross-linker for 30 min at room temperature. Reactions were quenched with 50 mM Tris, pH 7.5. The amount of protein in each sample was measured and normalized prior to gel loading. Samples were mixed with 8% 2ϫ SDS sample buffer and incubated for 30 min at room temperature before fractionation by a 4 -8% Tris-glycine PAGE and analysis by Western blotting.

Fluorescence Resonance Energy Transfer
Fluorescence resonance energy transfer (FRET), implemented on a Zeiss LSM 510 confocal microscope (Carl Zeiss Optics Company, Jena, Germany), was used to measure the degree of prestin self-association following cholesterol perturbations. HEK cells were cotransfected with prestin-CFP (donor) and prestin-YFP (acceptor). Both the CFP and YFP fusion proteins had an engineered mutation, A206K, that prevents CFP and YFP dimerization to allow easier interpretation of FRET results. Details of the acceptor photobleach technique utilized have been published previously (61). Briefly, a region of interest (ROI) on the cell membrane, exhibiting even, membrane localized fluorescence, was bleached to remove YFP signal. CFP fluorescence intensity in the bleached ROI was measured pre-and post-bleach to arrive at the value of FRET efficiency (E f ) from CFP to YFP. CFP intensity in an adjacent unbleached ROI was measured pre-and post-bleach to derive control (C f ) values for each FRET measurement. For detailed description of methods, refer to Greeson et al. (61).

Quantification of Puncta
Membrane segments used for FRET experimentation were also utilized in puncta quantification. Confocal images of prestin-YFP in living HEK 293 cell membranes were cropped to ϳ25 by 75-pixel (1 pixel ϭ 0.14 m) membrane-containing regions (slice thicknesses ϭ 3.3 m) and analyzed using the Matlab (The Mathworks, Natick, MA) edge detection filter. The filter generated an image the same size as the input image composed of ones where edges were detected and zeros everywhere else (supplemental Fig. 3). Edge detection is based on one of six specific methods, and the most advanced of the available methods, the Canny method, was used in this work. The Canny method identifies local maxima in the image gradients and uses two different threshold values to define both strong and weak edges. A weak edge is only included in the output image if it is connected to a strong edge. The use of two threshold values ensures that this method is less likely to detect false positive weak edges. Following application of the filter, the output image was analyzed for the presence of puncta. Puncta identification was guided by the output image of the detection filter and supported by the membrane region image (supplemental Fig. 3). The results of this analysis are shown in Fig. 5. The same traces, on a magnified y axis, are shown in the inset. DPOAE recordings during depletion showed higher noise levels (2.4 db S.D. from the average trend line) post-delivery, when compared with loading (0.34 db) and control treatments (0.2 db), or to intrinsic noise in the recordings before delivery (0.42, 0.89, and 0.31 for control, depletion, and loading treatments, respectively). Arrowheads indicate times at which the round window was perforated to eliminate DPOAEs.

Statistical Analysis
V pkcs and capacitance gains in OHCs and prestin-transfected HEK 293 cells subjected to different treatments (untreated, cholesterol loading, and cholesterol depletion) were compared using two-tailed t tests. Statistical significance (in comparison to untreated) is indicated in Table 1. Two-way analysis of variance was also used to evaluate statistical significance of FRET values in comparison to control FRET for each treatment, as well as to FRET values of untreated cells. Statistical significance is indicated by asterisks in Fig. 7D and Table 1.

Changes in Cholesterol Alter the Amplitude of Distortion
Product Otoacoustic Emissions (DPOAEs)-To study the effects of cholesterol on hearing at the organ level, we evaluated cochlear function in vivo by measuring DPOAE amplitudes during cholesterol alteration (Fig. 1). Cholesterol depletion resulted in a 20-db decrease in DPOAE amplitudes. Cholesterol loading, on the other hand, resulted in an initial 2-3-db increase in DPOAE amplitude, followed by a decrease of up to 20 db (Fig.  1). These results confirm that cholesterol modulates hearing.
Because DPOAE amplitudes reflect OHC electromotility, the level of cholesterol in the OHC lateral wall may be functionally relevant.
OHC Lateral Wall Cholesterol Content Decreases with Maturation-To visualize cholesterol in the OHC lateral wall, we used filipin labeling of mouse OHCs (Fig. 2). Filipin labeling clearly indicates membrane cholesterol content, as shown by filipin labeling of untreated, cholesterol-depleted and -loaded HEK 293 cells (supplemental Fig. 1). OHCs from P6 and P12 mice showed distinct filipin labeling of the lateral wall, compared with cytoplasm. However, OHCs from adult mice showed intracellular filipin staining surrounded by a region of lower staining corresponding to the membrane (Fig. 2, right  panels). The filipin staining patterns are clearly visible in the pixel intensity graphs (Fig. 2, bottom row), which plot pixel intensities along the diameter of a single OHC in each case. These data indicate that the cholesterol in the OHC lateral wall is initially high and decreases during development. This time frame parallels the onset and maturation of OHC electromotility (65), which includes a depolarizing shift in prestin-associated charge movement.

Alterations in Cholesterol Content Modulate Nonlinear Capacitance in Isolated
OHCs-We directly evaluated the effect of cholesterol on prestin-associated charge movement at the cellular level by altering cholesterol content in isolated guinea pig OHCs. We observed characteristic bell-shaped voltage-dependent capacitance (NLC) with a peak, V pkc , at about Ϫ0.050 V in untreated OHCs (Fig. 3A), similar to that reported earlier (66). Depletion of cholesterol shifted V pkc in the depolarizing direction to about ϩ0.080 V, whereas loading excess cholesterol shifted V pkc toward more hyperpolarizing voltages (less than Ϫ0.130 V, see Fig. 3A). Kinetic studies of these phenomena indicated that the shift in V pkc occurred within minutes of adding water-soluble cholesterol (loading) or M␤CD (depletion) (Fig. 3B). Furthermore, the effects of depletion and loading were reversible, and the reversal of the V pkc shift was equally rapid (Fig. 3C). These data (statistics represented in Table 1) indicate a direct and dynamic correlation between OHC membrane cholesterol content and V pkc .
Changes in Membrane Cholesterol Reversibly Alter V pkc in Prestin-transfected HEK 293 Cells-We next investigated prestin-specific effects of cholesterol in HEK 293 cells. The V pkc of prestin-transfected HEK 293 cells was approximately Ϫ0.070 V (Fig. 4A). Upon cholesterol depletion, the V pkc shifted toward depolarized voltages, with an average peak at ϩ0.004 V (Fig.  4A). On the other hand, upon cholesterol loading, the V pkc shifted toward hyperpolarized voltages, with an average value of about Ϫ0.116 V (Fig. 4A). Importantly, these effects are reversible. Cholesterol depletion followed by loading, as well as cholesterol enrichment followed by depletion, both shifted the peak voltage toward the control untreated average (Fig. 4B). The kinetics of these processes were similar to those measured for OHCs, with changes in the V pkc occurring within minutes of addition of cholesterol or M␤CD (Fig. 4C), and the process was rapidly reversible (Fig. 4D). Comparisons of changes in V pkc in  Table 1.

TABLE 1 V pkc of nonlinear capacitance in OHCs and prestin-expressing HEK 293 cells
Mean values and S.D. are indicated. Sample sizes for each group are indicated in parentheses. Statistical significance of each group (in comparison to untreated cells of the same type) is also indicated: ** ϭ p Ͻ 0.0001.  (Fig. 5). Prestin colocalizes with the plasma membrane (Fig. 5A), consistent with earlier observations (59,60), and is expressed in foci suggestive of localization to membrane microdomains (Fig. 5, A and D). Alterations in cholesterol content change the distribution of the foci; upon cholesterol depletion (Fig. 5, B and E), prestin showed a less punctate (more uniform) membrane distribution, whereas upon cholesterol loading (Fig. 5, C and F), the number of foci increased (Fig. 5G). Quantification of puncta (Fig. 5G) was based on multiple images of membrane segments from different batches of treated HEK 293 cells (supplemental Fig. 3). In addition, time-lapse images of a single transfected cell taken during the course of depletion or loading show a clear decrease and increase in puncta (supplemental Fig. 4). These results demonstrate that cholesterol influences prestin distribution within the membrane.
To assay prestin localization in membrane microdomains, we characterized detergent-resistant membrane extracts isolated from prestin-transfected HEK 293 cells. As expected, prestin was detected in the dense endoplasmic reticulum membrane fractions (Fig. 6A, lanes 8 -10), which cofractionated with the endoplasmic reticulum markers (59). Prestin was also seen in the less dense plasma membrane fractions and predominantly localized in membrane microdomain fractions (Fig. 6A,  lanes 4 and 5), where it cofractionated with flotillin-1, a microdomain marker and structural component (67). Upon cholesterol depletion prestin, but not flotillin, redistributed out of the microdomain fractions (Fig. 6B). Upon cholesterol loading, prestin remained in microdomain fractions with higher intensities of all bands (Fig. 6C). These data indicate that prestin can localize to membrane microdomains and that cholesterol modulates its distribution in these domains.

Manipulation of Cholesterol Content Alters Prestin
Associations-Localization to microdomains raises the possibility that prestin may interact with itself or other proteins. To  Table 1. DECEMBER 14, 2007 • VOLUME 282 • NUMBER 50 determine whether prestin-prestin interactions were altered by cholesterol, we analyzed prestin complex associations by crosslinking (Fig. 7). In the absence of cross-linker (lane 1), a weak prestin dimer band exists in untreated cell extracts (Fig. 7A). This band disappears upon cholesterol depletion (Fig. 7B) and is stronger upon cholesterol loading (Fig. 7C), indicating cholesterol favors prestin self-association in HEK 293 cells independent of cross-linking. Prestin dimers were present at relatively low BS 3 cross-linker concentrations (Fig. 7A, lane 2), and the dimer band intensified with increasing cross-linker concentrations (Fig. 7A, lanes 3-7). Cholesterol depletion caused a decrease in prestin cross-linking, with higher concentrations of BS 3 required to trap prestin as dimers (Fig. 7, compare lanes 2  and 3 in B with lanes 2 and 3 in A). In contrast, loading cholesterol into the membrane caused the appearance of oligomeric protein bands even in the absence of cross-linker (Fig. 7C, lane   1) and bands intensified with increasing cross-linker (Fig. 7C,  lanes 2-8). In conjunction with PFO-polyacrylamide gels (supplemental Fig. 2A) and coimmunoprecipitation experiments (supplemental Fig. 2B), and in accordance with earlier observations (61,68,69), these data provide strong evidence of prestin self-association and indicate that prestin self-association increases with cholesterol content.

Cholesterol Modulates OHC Function
Further evidence of cholesteroldependent prestin self-association was obtained using acceptor photobleach FRET. We found an average FRET efficiency of 6.7% (Fig. 7D) in live, untreated HEK 293 cells cotransfected with prestin-CFP and prestin-YFP. Cholesterol enrichment increased average FRET efficiency to 8.4%, whereas cholesterol depletion resulted in a marked decrease of FRET efficiency to 0.5% (Fig. 7D), indicating significant loss of prestin self-association. In this case, the measured FRET efficiency is indistinguishable from corresponding control efficiency measurements (p Ͼ 0.05). To determine whether this elimination of prestin self-association is reversible, cells were first depleted of membrane cholesterol and then reloaded. Upon reloading, FRET efficiency returned to 6.8% (Fig. 7D). These results confirm that modulation of prestin self-association by membrane cholesterol is dynamic and reversible. Our organ, cellular, and molecular level studies greatly expand our understanding of the role of cholesterol in tuning the functionality of the outer hair cell motor.

DISCUSSION
Cholesterol is crucial for membrane organization and dynamics and in the regulation of membrane protein sorting and function. Considering that electromotility is based in the highly specialized and cholesterol-poor OHC lateral membrane, the established link between elevated cholesterol and auditory dysfunction (27,28,34,70,71) suggests direct effects of cholesterol on the OHC membrane and proteins present therein. Models for OHC electromotility based solely on the electromechanical transduction capabilities of the membrane have been proposed (72,73). Following the discovery of the membrane protein prestin, numerous studies have demonstrated how alterations of membrane material properties affect prestin function and/or OHC electromotility (67, 74 -79). These studies point to a dynamic interplay between prestin and the membrane in the generation of nonlinear capacitance and electromotility; our study further characterizes this dynamic relationship.
Cholesterol Effects on Otoacoustic Emissions Correlate to Effects on NLC-Cochlear cholesterol alterations influence DPOAE amplitudes (Fig. 1). Manipulating cholesterol in isolated guinea pig OHCs revealed a dynamic and reversible relationship between membrane cholesterol content and prestinassociated charge movement, with similar kinetics (Fig. 3). Cholesterol effects on DPOAEs may result from the observed effects on prestin-associated charge movement as follows. In normal untreated OHCs, the NLC peak is at approximately Ϫ0.050 V; therefore, in the cell's range of receptor potential (nominally between Ϫ0.060 and Ϫ0.080 V) capacitance is sub-maximal. Shifting the NLC peak in the depolarizing direction (as upon cholesterol depletion) would result in a progressive lowering of capacitance and electromotility in the range of the cell's receptor potential. This would reduce DPOAE amplitudes. On the other hand, shifting the NLC peak in the hyperpolarizing direction would result in an initial increase of capacitance at the cell's receptor potential followed by a decrease. Electromotility and DPOAE amplitudes would follow this pattern. This relationship is schematically presented in Fig. 8.
OHC Function May Be Modulated by Cholesterol Reduction during Maturation-We show a lowering of membrane cholesterol with maturation (Fig. 2). During the post-natal maturation of rodent OHCs, the distribution of prestin in the lateral wall is initially inhomogeneous (45,65). Concurrent with prestin distribution becoming homogeneous, maturation of electromotility and nonlinear capacitance is observed, which includes a shift in V pkc from an immature hyperpolarized value to the normal adult value (45,65). Both effects may result from a decrease in OHC membrane cholesterol levels with maturation. Our data suggest that the reduction in membrane cholesterol with maturation helps to tune the membrane-based motor to operate at maximal gain in the OHC receptor potential range.
HEK 293 Cells Provide a Model System for Studying Prestin Function-Cholesterol manipulations in prestin-transfected HEK 293 cells (Fig. 4) produced qualitatively similar results as in OHCs (Fig. 3; Table 1), validating the use of HEK 293 cells as a model system. The difference between V pkc in cholesteroldepleted OHCs and HEK 293 cells may be due to structural differences (membrane tension and turgor pressure) or differences in cholesterol homeostasis mechanisms, between the two cell types, which cause similar trends but different magnitudes in the effects of depletion.
Prestin-transfected HEK 293 cells allowed for histological and biochemical analyses following alterations in cholesterol. Prestin appears to be present in foci characteristic of membrane microdomains in HEK 293 cells (Figs. 5 and 6). Similar foci have not been observed in the adult OHC lateral wall membrane. Perturbing cholesterol content alters the distribution of prestin; prestin shifts out (cholesterol depletion, Fig. 6B) or remains in (cholesterol enrichment, Fig. 6C) the microdomain fractions, suggesting that prestin is capable of localizing to cholesterolrich microdomains. In addition, a quantitative correlation exists between cholesterol content and the number of prestin puncta in the membrane; cholesterol depletion results in a reduction, whereas enrichment causes an increase in number of foci (Fig. 5G).
Recent studies have suggested that prestin may self-associate and dimerize (61,68,69). We have obtained further evidence of prestin self-association using a cross-linking reagent, which revealed the presence of prestin-prestin interactions that are decreased upon cholesterol depletion and increased upon cholesterol addition (Fig. 7, A-C). Furthermore, FRET measurements provide direct evidence of the significant and reversible effect of cholesterol on prestin self-associations (Fig. 7D). In light of these data, the low cholesterol levels on the mature OHC lateral wall are consistent with the homogeneous distribution of prestin.
Mechanism of Cholesterol Effects-The effects of cholesterol on membrane protein function have been the subject of numerous recent studies, and several mechanisms have been put forth to explain the effects of cholesterol. In addition to the effects on membrane material properties such as viscosity, elasticity, compressibility, and stiffness (80), cholesterol levels in the membrane influence the formation of ordered microdomains (81,82) and partitioning of proteins into these domains by altering the bending modulus of the membrane (16) and thereby influencing hydrophobic mismatch (15). The same fac-  4 and 5). B, depletion of cholesterol with 10 mM M␤CD causes a redistribution of HA-prestin into heavier membrane fractions. C, cholesterol enrichment (10 mM water-soluble cholesterol) enhances colocalization into membrane microdomain fractions. The arrowhead and black arrow point to unglycosylated and glycosylated monomeric prestin, respectively. The white arrow points to oligomeric species. Shown are representative data from a single experiment; the experiment was repeated at least three times. DECEMBER 14, 2007 • VOLUME 282 • NUMBER 50 tors may explain the effect of cholesterol content on prestinassociated charge movement. Because cholesterol is known to modulate membrane material properties, which in turn affect the dynamics of membrane proteins, cholesterol-dependent changes in membrane stiffness or curvature could alter the dynamic fluctuations of prestin as would changes in membrane dipole potential and lipid packing density (19). Several studies in the OHC have correlated changes in membrane tension, stiffness, and mechanics to changes in the V pkc and electromotility (67,(75)(76)(77)79). Molecular dynamics simulations suggest cholesterol affects lipid lateral pressure profiles (17,18), and this would impact the prestin conformational change that is assumed to accompany charge movement.

Cholesterol Modulates OHC Function
Cholesterol-induced V pkc shifts in both OHCs and HEK 293 cells are larger than those resulting from previous manipulations; these include exogenous chlorpromazine (77), fructose (46), and increasing intracellular pressure (76,83), which shift V pkc toward depolarizing potentials; and decreasing intracellular pressure and exposure to the lipophilic ion, tetraphenylborate (TPB Ϫ ) (84), which move V pkc in the hyperpolarizing  1-8, respectively) of the membrane-impermeable agent BS 3 . B, cholesterol depletion using 10 mM M␤CD causes a reduction in cross-linking; higher concentrations of BS 3 are required for dimer formation. C, cholesterol loading using 10 mM water-soluble cholesterol causes an increase in cross-linking; oligomer bands appear even in the absence of cross-linker (lane 1). M, D, and T denote monomeric, dimeric, and trimeric prestin bands, respectively (based on molecular weight). D, acceptor photobleach FRET measurements to evaluate prestin self-association in live HEK 293 cells. Acceptor photobleach FRET efficiencies (f) and control (unbleached) FRET values (f) were measured from untreated (n ϭ 22), cholesterol-depleted (n ϭ 20), cholesterol-loaded (n ϭ 16), and depleted and reloaded (n ϭ 23) prestin-expressing HEK cells. Statistical significance (in comparison to control FRET for each treatment) is represented; *, p Ͻ 0.05. Shown are representative data from a single experiment; the experiment was repeated at least three times. FIGURE 8. Schematic representation of correlation between NLC peak shifts and electromotility. The nonlinear capacitance of untreated OHCs has a peak at about Ϫ0.050 V, slightly depolarized from the resting potential of the cell. The corresponding capacitance in the operating range (receptor potential) of the cell (indicated by gray box and sinusoidal wave) is therefore slightly sub-maximal. Upon depletion, the peak shifts further away from this operating range, resulting in a progressive reduction in capacitance in this range. Upon loading, the peak initially shifts into the operating range, resulting in small increase in capacitance, and then shifts beyond the operating range resulting in a decrease of capacitance in the range. Electromotility and otoacoustic emissions may be presumed to follow the same pattern.
direction. Because the magnitude of the V pkc shifts is significantly greater than in previous manipulations that are known to change the material properties of the membrane, we must consider the possibility that the V pkc is also a function of self-association. This contribution would reflect the relative amounts of monomers versus higher order oligomers, where the monomeric form shifts V pkc to depolarizing voltages, whereas higher order oligomers shifts V pkc to hyperpolarizing voltages. The effect of cholesterol on both prestin self-association and on prestin function is reversible, indicating a dynamic interaction of prestin with membrane components.
Cholesterol also has a propensity to localize to membrane microdomains. Prestin is present in microdomains in HEK 293 cells, and its presence in these localized domains may facilitate its interaction with itself or with other proteins. It is likely that prestin exists in a dynamic equilibrium between monomeric, dimeric, and perhaps higher order oligomeric forms. The effect of cholesterol might be to "cluster" prestin molecules, shifting the equilibrium toward dimeric or oligomeric species. Our data indicate increased self-association in the presence of increased cholesterol. The low cholesterol level observed in the mature OHC lateral wall suggests a preference for lowered prestin selfassociation, the functional consequences of which remain to be studied.
In summary, our study integrates systems-level, cellular and molecular data to investigate the role of cholesterol in modulating the mechanical aspects of mammalian hearing. We have characterized interrelationships between prestin-prestin interactions and prestin-membrane interactions. Whether the effect of cholesterol is predominantly through formation of functionally distinct microdomains, changes in membrane material properties, or both, the observable effects of changing the cholesterol content are a change in prestin self-association, a reversible shift in V pkc , and changes in otoacoustic emissions. This reinforces the concept of the molecular motor driving electromotility as an interdependent entity with protein and membrane components working cooperatively to achieve nonlinear charge movement and mechanical motion.