Organization of membrane motor in outer hair cells: an atomic force microscopic study
- First Online:
- Cite this article as:
- Sinha, G.P., Sabri, F., Dimitriadis, E.K. et al. Pflugers Arch - Eur J Physiol (2010) 459: 427. doi:10.1007/s00424-009-0742-3
- 376 Downloads
Using atomic force microscopy, we imaged the cytosolic surface of the lateral plasma membrane of outer hair cells from guinea pigs’ inner ear. We used a “cell-free” preparation, in which a patch of plasma membrane was firmly attached to a substrate and the cytoplasmic face was exposed. The membrane patches contained densely packed particles whose diameter, after correcting for the geometry of the probing tip, was ∼10 nm. The particles were predominantly aligned unidirectionally with spacing of ∼36 nm. The density of the particle was ∼850 μm−2, which could be an underestimate presumably due to the method of sample preparation. Antibody-labeled specimens showed particles more elevated than unlabeled preparation indicative of primary and secondary antibody complexes. The corrected diameters of these particles labeled with anti-actin were ∼12 nm while that with antiprestin were ∼8 nm. The alignment pattern in antiprestin-labeled specimens resembled that of the unlabeled preparation. Specimens labeled with actin antibodies did not show such alignment. We interpret that the particles observed in the unlabeled membranes correspond to the 10-nm particles reported by electron microscopy and that these particles contain prestin, a member of the SLC26 family, which is essential for electromotility.
KeywordsOuter hair cellsAtomic force microscopyPrestinCytosolic plasma membraneHair CellMicroscopyMechanoelectrical transductionImmunocytochemistry
Electromotility [2, 7, 28] is a conspicuous motile property of cochlear outer hair cells (OHCs) and is driven by electrical energy available at the plasma membrane. Along with hair bundle motility [8, 30], this motility of the cell body is capable of producing fast mechanical force in response to mechanical stimuli and thereby amplifying the mechanical signal. It is therefore a critical factor for the sensitivity, frequency selectivity, and nonlinear characteristics of the mammalian ear [9, 35].
Electrophysiological studies showed that motile activity is associated with charge transfer across the membrane [3, 43], similar to piezoelectricity [14, 19, 27, 38]. The mobile charge contributes to the membrane capacitance as a nonlinear, voltage-dependent component. According to two-state models, the steepness of the voltage dependence indicates a unit mobile charge of about 0.9 e and the ratio of the total charge transfer to the unit charge gives a motor unit density between 5,000 and 10,000 μm−2 in OHCs from the middle turns of the cochlea [29, 43, 45]. The transfer of this unit mobile charge is coupled with area change of about 4 nm2 .
A molecular biological approach identified prestin, a member of the SLC26 family of anion transporters, as the protein associated with the motile activity of the lateral wall. Cells transfected with prestin show nonlinear capacitance, characteristic of OHCs. The steepness of the voltage dependence is about the same as OHCs [41, 57], and the corresponding area change is about 2 nm2 . As is often the case with membrane proteins, prestin appears to form oligomers [40, 58], presumably about 10 nm in diameter [37, 39].
Morphological studies with electron microscopy showed that the lateral plasma membrane is packed with particles of about 10 nm in diameter (the “10-nm particles”) [17, 21, 29, 45]. These particles are not apparent from the extracellular face with freeze fracture method without etching . The reported density of those particles ranges between 2,500 and 6,000 μm−2 [17, 29, 45]. Although the density of the 10-nm particles is similar to that of the motor molecule determined from electrophysiological experiments in the order of magnitude, the relationship between the 10-nm particle and the functional unit remains unclear because their ratio shows significant uncertainty and variability .
An important question that has not been directly addressed, even though assumed in some of the studies [37, 39], is the molecular identity of the 10-nm particles, specifically whether or not prestin is a significant constituent of the 10-nm particles. Prestin forms oligomers, which are consistent with the dimension of the 10-nm particles, in the plasma membrane of prestin-transfected cells. However, there is no direct proof that prestin oligomers are indeed 10-nm particles, because the lateral plasma membrane of OHCs and the plasma membranes of cultured cells should contain membrane proteins other than prestin. In addition, different cytoskeletal undercoatings of these membranes could affect the organization of the membrane proteins. Specifically, the undercoating of the lateral plasma membrane of OHCs is elaborate and unique . For those reasons, the 10-nm particles in OHCs and oligomers in the plasma membrane of prestin transfected cells may not be identical. In addition, the issue of oligomerization of prestin itself may not be very simple. Although studies using gel electrophoresis show the involvement of disulfide bonds in dimerization [52, 58], suggesting relative stability of oligomers, studies that use fluorescence resonance energy transfer (FRET) show that dimerization of prestin is sensitive to cholesterol content in the plasma membrane and that the FRET signal increases or decreases with cholesterol content of intact OHCs . This observation suggests that prestin monomers may exist in the plasma membrane of OHCs. Those reports therefore pose an additional question as to whether prestin exists outside of the 10-nm particles even if the 10-nm particles contain prestin.
In this study, we specifically probe the molecular identity of these 10-nm particles by using atomic force microscopy (AFM). To obtain the spatial resolution required, the extracellular side of the plasma membrane must adhere firmly to a flat substrate and the cytosolic side must be exposed. Although an earlier report shows an AFM image of such a patch of the plasma membrane that is broken off from a fixed outer hair cell , the image is too complex to analyze, apparently being dominated by intracellular structure such as the cortical cytoskeleton. Thus, it seems necessary to tear off a patch of membrane before fixing the cell. Such a method, called “cell-free” preparation, has been successfully used for tissue-cultured cells [4, 23, 39]. Here, we apply this method on acutely dissociated outer hair cells and image the membrane patch with AFM.
We also tested the immunoreactivity with prestin antibodies by examining the sizes and organization of the molecular complexes that these antibodies form. We show that imaging the bound antibodies at the single molecule scale with the atomic force microscope, a relatively unexplored method, has the advantage of viewing the arrangement of the antigen–antibody complexes when compared with the current optical methods in use for immunohistochemistry.
Materials and methods
Outer hair cells were dissociated from the organ of Corti of pigmented guinea pigs (Elm Hill Labs, Chelmsford, MA, USA) between 200 and 400 g in body mass following a method previously described  and approved as protocol 1195-05 NINDS/NIDCD. Briefly, the organ of Corti was dissected from the cochlea in a perilymph-like medium, which contained 140 mM NaCl, 5 mM KCl, 2 mM MgCl2, 1.5 mM CaCl2, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 2 mM Na pyruvate, 2 mM creatine, and 2 mM Na ascorbate with pH adjusted to 7.3 with NaOH, which we refer to as the standard medium. The Hensen cells were then removed with a needle. The tissue obtained was then treated with 1 mg/ml dispase (Worthington) for 15 min at 23°C before being gently triturated three times with a plastic pipette. The isolated cells were then placed in a chamber created on an aldehyde-coated glass slide (superaldehyde substrates, ArrayIt Microarray Technology, Sunnyvale, CA, USA), to which a thin plastic film with a printed grid, with 1 mm spacings, was glued to the slide underside. The chamber was centrifuged at 200×g for 20 min. The positions of isolated cells with good optical density were recorded with reference to the grid so that these locations could be examined later for isolated membrane patches (cell-free preparation). This preparation was then briefly (for about 0.1 s) sonicated using an ultrasonic cell disrupter (XL-2000, Misonix, Farmingdale, NY, USA) at a low power setting to shear open the cells. The chamber was then carefully rinsed with the standard medium to remove floating cellular debris.
To obtain images of unlabeled specimens, the prepared samples as described above were immediately treated with 2.5% glutaraldehyde for 10 min. They were then washed with the standard medium and placed on the AFM Bioscope stage for scanning.
To examine the immunoreactivity of the preparation, we used prestin antibodies (provided by Dr. B. Kachar, NIDCD, NIH) that were raised in rabbits against a peptide from the C terminus domain of gerbil prestin (amino acids 725–744) . The sonicated and rinsed samples were treated with antiprestin primary antibodies at ∼5 μg/ml concentration for 30 min . Then, the samples were washed three times before being treated with fluorescein-conjugated secondary antibodies (Amersham Pharmacia Biotech, Arlington Heights, IL, USA) at a 1:200 dilution for 20 min. These samples were rinsed three times before fixation with 2.5% glutaraldehyde. This was followed by several rinses, and then the samples were scanned with AFM. Antibodies were tested for their specific immunoreactivity on the organ of Corti by immunofluorescence labeling before use.
Anti-actin and anti-DNase II-labeled specimens
Actin antibodies (A5060, Sigma-Aldrich) and DNase II antibodies (D1689, Sigma-Aldrich) raised in rabbit were used as control antibodies. The protocol was the same as the one for antiprestin-labeled specimens except that ∼20 μg/ml actin antibodies or 0.25 μg/ml DNase II antibodies were used instead of prestin antibodies.
Atomic force microscopy
Bioscope AFM (Veeco, Santa Barbara, CA, USA) mounted on an inverted fluorescence microscope (Eclipse TE300, Nikon) was used with a Quadrexed Nanoscope IIIa controller. All the measurements were done in fluid tapping mode with V-shaped cantilevers mounted on a fluid holder. The standard medium was used as imaging buffer. In most cases, we used Si-N cantilevers, 200 μm in length, with a spring constant ≈0.06 N/m and oxide sharpened tips (DNP-S, Veeco, Santa Barbara, CA, USA). The feedback gains and scan speed were optimized before scanning. The typical amplitude of tip vibration was set between 0.5 to 1 V at scan speeds of ∼0.8 Hz per pair of scan lines. The tapping frequency was about 8 kHz, slightly lower than the cantilever resonance frequency, such that the amplitude of oscillation at the selected frequency was 5–10% below the amplitude at resonance frequency. All scans shown were performed from left to right, where the height profiles and phase contrast images were recorded simultaneously. Height profiles revealed the topography of the scanned surface, whereas phase contrast images originated from the phase lag of the oscillating tip following differences in height and variations in the probe sample interaction.
Data acquisition and analysis
To reduce noise in the autocorrelation function of images, we preprocessed images using two methods. One is background subtraction. Images processed with instrument software were converted to gray-scale images before imported to ImageJ or Matlab. A background image of each image was then created with Gaussian blurring, and then it was subtracted from its original image using ImageJ (Wayne Rasband, NIH).
The vectors x and x′ represent the horizontal plane, and p and i are functions at these points, which represent the inverted probe surface and image surface respectively. pj is the surface of the probe obtained after j iterations starting from initial probe surface p0. The extremum operation is implied to extend over all in-plane coordinates of the respective domains. DI is the domain of the image surface I. DP′ is the domain of P′ determined by considering all possible touch points, which have the apex on or below the image surface. d is the displacement vector between the touch points of the tip on the image and tip apex. The maximum operation within curly bracket is a union of the corresponding revised tip shapes for all possible d’s. The final tip is obtained from the best fit of the probes over the whole image as given by the minima operation over the domain of I. In this paper, a commercially implemented blind reconstruction procedure (SPIP, Image Metrology A/S, Denmark) was used to obtain the radius of the sphere equivalent to the tip curvature of the probe (sphere fit) together with the angle of the cone, which represents the gross shape of the probe. The images on which the procedure was applied were beforehand flattened and Gaussian filtered (5 × 5 kernel, cut off wavelength 5 nm) to remove noise contributions that may interfere with good tip estimation.
To our knowledge, cell-free preparations so far reported were formed from tissue-cultured cells [4, 23, 39]. Our preparation differed from those reports in that it was made from acutely dissociated cells, which were not homogeneous. Our method of isolating outer hair cells could not completely remove Deiters’ cells or pillar cells. However, these supporting cells had turned into ghosts because their plasma membranes had broken during isolation, being firmly attached to the basilar membrane. Only outer hair cells retained unbroken plasma membranes. The percentage of outer hair cells forming strong bonds to the substrate during centrifugation is thus expected to be very high. The origin of isolated membranes in the cell-free preparation was identified by the location of the cells prior to sonication (see “Materials and methods” section). Such “cell-free” candidates are difficult to identify by standard fluorescent tagging techniques due to the short path length of the light source through the thin membrane. The membrane patches were selected out of several AFM scans of the premarked positions, where a relatively smooth elevated structure could be observed. Images with fiber like structure that could be part of the cortical cytoskeleton were excluded. Eleven unlabeled patches were obtained from different experiments, and a total of five were analyzed for particle size, height, and organization.
To determine the dimension of the particles observed, we first plotted profiles of sections, which cut across several particles of similar sizes (Fig. 2c, e) on flatter regions of the membrane patch. The apparent diameter of the particles obtained from a height image was 21.4 ± 3.3 nm (mean ± SD, N = 168). The apparent diameter of those particles obtained from the phase image, defined as the distance between their steepest slope points in this case, was 22.0 ± 2.4 nm (mean ± SD, N = 137; Fig. 2e).
The reconstructed tip from the image (see “Materials and methods” section) had a large cone angle of 156° presumably reflecting long-range nature of the interactions the probe has with the sample in the aqueous medium. Particle density as determined by granulometric filters of ImageJ (see “Materials and methods” section) was ∼850 μm−2. Given the large angle of the effective cone of the probe, it is likely that only those particles at the top are imaged, possibly underestimating the number of particles in the membrane. These imaged membrane particles appeared to be aligned (Fig. 2a, b). To study the distribution pattern, we calculated an autocorrelation function (see “Materials and methods” section), shown in Fig. 2 h, i. There were two prominent peaks that are about 36 nm away from the center. The line connecting these two peaks agreed with the orientation of the alignment that we can detect by direct visual inspection of the image. The autocorrelation function confirmed linear alignment and indicated the main axis and mean spacing of 35.5 ± 1.2 nm (N = 16). This alignment is unlikely a scanning artifact because it is at approximately 45° to the scanning axis. In addition to these peaks, there were less prominent peaks, a pair at an angle of 64° from the main axis and at 48 nm from the center and another pair at a 98° angle and a distance of 56 nm from the center. These three axes may be reminiscent of a hexagonal array. However, the angles that these axes formed deviated from 60°. In addition, peaks in the autocorrelation function were present only in the immediate proximity of the central peak. This differs from the autocorrelation function of a crystalline lattice, in which multiple peaks are regularly distributed. This difference indicates that the membrane particles that we observed had only short-range correlation and that they distributed rather randomly beyond their immediate neighbors.
Because prestin is the key protein for electromotility of the lateral membrane, we tested the immunoreactivity of our preparation using antibodies raised against polypeptides of the C-terminal region of prestin. To observe the geometry of the antibody complex formed as well as control on our preparation, we used actin antibody. The primary antibodies were conjugated to secondary antibodies to make them more readily recognizable (see “Materials and methods” section).
Six patches were identified from all the prestin-labeled preparations out of which four were selected for further analysis. One out of three patches was analyzed for the anti-actin sample while one out of three was analyzed for anti-DNASE II sample.
The membrane particles were aligned in one direction (Fig. 4a). The spacing between the particles of these prestin-labeled specimen was 38.7 ± 2.7 nm (N = 16), which was slightly larger than the spacing for the unlabeled preparation. However, unlike the unlabeled specimens, alignment in other directions was not clear by visual inspection. These images were not scanning artifacts because scanning the same area in different directions also produced similar results. This observation was confirmed by the autocorrelation function (Fig. 4f, g). The density of those particles was about 550 μm−2, lower compared to the 850 μm−2 for the 10-nm particles in the unlabeled specimens. The low value for the antibody density could be due to steric hindrance between antibody complexes. The preservation of the main alignment pattern of 10-nm particles confirms specific binding of prestin antibodies to 10-nm particles. With possible strain by antibody binding and lower density, the disappearance of weaker alignments may not be surprising. There are two alternative interpretations of the particles imaged in this preparation. One is that the 10-nm particles are made larger by the binding of antibody complexes and another is that the particles imaged are primary and secondary antibody complexes. The issue of interpretations will be addressed in “Discussion” section.
The blind reconstruction method gave an overall average tip radius of 19.4 ± 6.3 nm (N = 10). The value obtained from three consecutive images shown in Fig. 4e is 12.5 ± 2.5 nm. If we can assume that the imaged particles are antibody complexes that are attached to the surface of the plasma membrane, then h = 0 in Eq. 4. This interpretation leads to 8.2 ± 2.4 nm as an estimate for the particle diameter using 12.5 nm for the radius of the tip.
The number density was about 400 μm−2, lower than the 850-μm−2 density obtained for the unlabeled preparation and 550 μm−2 obtained for the prestin-labeled preparation. The distribution of those particles appeared random (Fig. 5a, b) and markedly different from the unlabeled specimens. This observation was also confirmed by the autocorrelation function (Fig. 5 g), where there is no clear peak other than the one in the center. Thus, the distribution of these particles did not reflect the alignment of the 10-nm particles. The randomness of the particle distribution cannot be due to the lower density. To prove this, we “reconstructed” (see “Materials and methods” section) an image on the basis of the particle locations of the antiprestin sample in which some of the particles were removed randomly to bring down the density comparable to the anti-actin sample. In such a “reconstructed” image, the autocorrelation still showed the alignment pattern similar to that of antiprestin sample.
The distribution of particles in the anti-DNase II-treated specimens was also random and did not reproduce the alignment pattern of the 10-nm particles in the unlabeled preparation (not shown). Moreover, this distribution was similar to that in the anti-actin-labeled preparation. These observations confirm that the random distribution pattern of particles in anti-actin and anti-DNase II preparations indicates nonspecific interactions, and thus, these preparations serve as controls.
Unlike immunofluorescence, we found that nonspecific site blockers were not useful in our examination using AFM. BSA, for example, is unsuitable because it has the size comparable to membrane proteins. Instead, not using the nonspecific blockers had the advantage of actually observing the particles, presumably antibody complexes, and their positional organization at nanometer scale.
Our observations support the hypothesis that the 10-nm particles contain prestin because the antiprestin-labeled specimens retain features including alignment and spatial periodicity of the unlabeled specimens. These features are not shared by the anti-actin-labeled specimens. This contrast shows that the alignment that is exhibited by antibody-untreated preparation is based on prestin and is unrelated to the cytoskeleton. In addition, this result is consistent with the assumption that prestin is located in a regular manner in the 10-nm particles and not in the plasma membrane outside of the 10-nm particles because the binding of antiprestin outside of the 10-nm particles could significantly disturb the alignment pattern of the unlabeled preparation. This point is highly relevant to the issue of prestin’s oligomerization. Recent studies with gel electrophoresis and FRET on prestin-transfected cells indicate oligomerization, most likely tetramerization of prestin [39, 40, 58]. However, oligomerization appears rather labile because removal and addition of cholesterol respectively reduces and increases oligomerization and the voltage sensitivity of prestin does not require oligomerization . Certainly, tetramer formation appears consistent with the particle morphology indicated by CryoEM  and AFM . How about the monomeric form? The existence of prestin monomers in the membrane of prestin-transfected cells can explain the FRET signal’s sensitivity to cholesterol content . Our observation does not suggest the presence of prestin outside of 10-nm particles in our preparation. It is possible that antiprestin could not bind to monomeric prestin due to steric hindrance of 10-nm particles. If that possibility can be excluded, virtually all prestin molecules are in the 10-nm particles in the plasma membrane of OHCs.
The unlabeled preparation shows a preferential axis with 35.5 ± 1.2-nm spacing. The antiprestin-labeled specimens show one alignment axis with 38.7 ± 2.7-nm spacing. Some stretching in the prominent axis and the disappearance of the less 11 prominent axis can be interpreted as result of disturbance due to antibody binding. Antibody binding may preserve regularity but may introduce some randomness, presumably eliminating weak alignment. The uneven structure may partly be causing minor differences in the alignment pattern. Roughness analysis showed that the maximum height range (Rmax) is 20.0 nm for antiprestin-labeled preparation and 36.6 nm for anti-actin-labeled preparation.
Here, we consider two possible models of antibody binding. One is that (1) an antibody complex, including a primary antibody and secondary antibodies, spreads out relatively thin when it attaches to a surface and another is that (2) it takes a form that can be approximated by a sphere. The first model is based on the reports that the AFM image of a single antibody molecule attached to a substrate has an apparent diameter of 24 nm and a height of 2–3 nm  and that binding of antibody increases the height of an IP3 receptor by 2–3 nm . The second model predicts a diameter of 9 nm based on the molecular weight of an antibody complex of ∼300 kDa and the density of ∼1.35 g/cm3 [20, 31, 49].
Our anti-actin-labeled preparation favored the second model. The elevation of the imaged particles is about 8 nm, inconsistent with the first model and consistent with the second model. In addition, the estimated diameter of 11.8 ± 4.7 assuming the second model is consistent with the expected diameter. If we assume that actin antibodies bind to 10-nm particles randomly in a manner consistent with the first model, the alignment pattern of the unlabeled preparation should be expected as it did in the antiprestin-labeled preparation. However, that is not the case.
Our antiprestin-labeled preparation, however, did not allow us to exclude either model. The elevation of 3 nm above the plasma membrane is consistent with the first model. Although this elevation may not be consistent with the second model, the estimated diameter of 8.2 ± 2.4 nm for the antiprestin complex based on the second model is still consistent with this interpretation. It could be speculated based on the second model that prestin antibodies might be bound to a particular site on the 10-nm particle presumably close to the membrane surface and that actin antibodies, which bind to the plasma membrane randomly, are more elevated.
It is likely that the AFM image of antibody complexes may depend on the surface to which they bind. Our observation on anti-actin-labeled preparation excludes the first model whereas previous reports [22, 51], which appear to have strong effects of glass and mica surface on antibody binding, exclude the second model.
Ten-nanometer particles and their alignment in the membrane
The particles in our unlabeled preparation have a horizontal diameter of 10 nm after correction. This value is consistent with the 10-nm particles reported by EM studies [29, 45]. The height of these particles is about 12 nm from the substrate, which is consistent with a recent EM study . This neural network analysis of EM images also show that these particles must protrude into the cytosolic side of the membrane. Such structure is consistent with the substantial cytosolic moiety, particularly near the C terminus of prestin, predicted by several reports [10, 40, 57]. This is also consistent with a freeze-fracture image showing dense particles protruding on the cytosolic face of the plasma membrane (B. Kachar, personal communication).
However, the value of 850 μm−2 for the density of the 10-nm particles in our preparation is lower than those obtained from EM studies, which range from 2,500  to 6,000 μm−2 . It is likely that 6,000 μm−2 is an overestimate because it includes particles of two different sizes . Those values are somewhat lower than the density, typically between 7,000 and 10,000 μm−2, of functional motor unit determined from nonlinear capacitance, assuming a two-state model [29, 45]. A lower bound of density estimates based on nonlinear capacitance data would be ∼2,000 μm−2 assuming that the functional unit is prestin monomer and the 10-nm particles are tetramers of prestin. Our lower value could be due to undulation of the sample surface. The value for the maximum height range (Rmax), 23.9 nm, obtained from roughness analysis is two times greater than the height of the membrane particles, in the unlabeled preparation. Undulation of the sample may have prevented the steeply tapered tip geometry (Fig. 4e) from reaching (and thus imaging) all membrane particles. However, this effect does not appear to be significant for the unlabeled preparation because a hexagonal packing with 35.5-nm spacing leads to the density of 920 μm−2, not significantly larger than the observed density of 850 μm−2. Thus, the lower density of 10-nm particle is more likely due to their dislodging during sonication or the effect of the medium we used, which is high in Cl− and Ca2+ concentrations.
The effect of the steeply tapered tip appears to have more significant effect on imaging the antibody-labeled preparation where the particle density of 550 μm−2 has a larger difference from the density of 770 μm−2 obtained by assuming a hexagonal packing with 38.7-nm spacing.
The spacing of 36 nm between the 10-nm particles in the unlabeled preparation is reminiscent of the half-period of the actin double helix . However, the structure observed cannot be due to the presence of outer hair cell cytoskeleton because labeling with actin antibody does not show such regularity. In addition, we do not see any resemblance to the structure observed in the AFM studies of outer hair cell cytoskeleton either by AFM  or by electron microscopy . These studies show that the OHC cytoskeleton comprises of parallel actin filaments 30 to 80 nm apart cross-linked by spectrin. The actin–spectrin network is connected to the plasma membrane via 25-nm-long pillars of unknown composition defining a height of at least 25 nm for the patch from a substrate. A recent publication also showed that cytoskeleton elements do not directly interact with prestin . The heights of our membrane patches are different from these values suggesting that our membrane patches do not contain structure in our preparation.
In our preparation, the 10-nm particles showed a distinct alignment pattern. In comparison, EM studies are not unanimous regarding the alignment of the 10-nm particles [17, 29, 45]. Random distribution is observed in glutaraldehyde-treated preparations . However, glutaraldehyde does not necessarily induce randomness because hexagonal patterns are observed in freeze fracture preparations with  or without  glutaraldehyde fixation. We compare our results with the freeze fracture EM images (without fixation) .
The EM images show semicrystalline regularity and equally strong alignment in two directions, which form a 68° angle. Although our images have symmetry axes similar to those preparations, alignment is more significant along one axis, and the correlation length (the extent of the ordered regions) is much shorter. Large membrane undulation and sonication uprooting might partly be responsible for the short correlation length. One possible factor responsible for the difference in the alignment pattern could be glutaraldehyde treatment. Another factor could be the Cl−- and Ca2+-rich medium that we used. However, perhaps the most likely factor for the difference in the alignment pattern would be the difference in the density of the 10-nm particles.
What is the origin of the alignment observed? It has been shown that anisotropy of the lateral wall in OHCs is based on the cortical cytoskeleton [17, 24, 25]. It would be possible that the cytoskeleton and membrane proteins associated with it could orient membrane proteins if the interactions between them are solid like. As we have discussed, our result obtained with actin antibodies revealed that the alignment pattern is not based on actin, the primary component of the cytoskeleton. We did not detect in our preparation a regular pattern that matches the description of “pillars” that are thought to connect the plasma membrane with actin filaments [17, 25]. In addition, solid-like interaction involving prestin is very unlikely because prestin diffuses easily along the plasma membrane . The diffusion coefficient of prestin is in the range 0.08–0.35 μm2 s−1 , much higher than the value 0.0001 μm2 s−1 for Na+ channels in skeletal muscle . For those reasons, the observed alignment may reflect symmetries in the internal organization of these particles. For example, hexagonal packing of purple membrane seems to reflect trimer formation of bacteriorhodopsin [18, 46, 50]. The square lattice of aquaporin corresponds to the molecule’s tetramerization .
We made an attempt at explaining the particle alignment that we observed, assuming that each 10-nm particle is a homologous tetramer of prestin molecules to illustrate a possible internal symmetry of these particles. The reason for this assumption is that it is the simplest. It has been pointed out that homologous oligomerization may lead to a reduction in voltage sensitivity of electromotility due to increased charge interaction . Heterologous oligomers may not suffer such an effect. Indeed, we do not have any reason to reject the possibility of heterologous oligomers. Currently, however, we have no data to construct any heterologous model. A homologous tetramer is consistent with previous reports [11, 36, 37, 39, 40, 42, 47, 58] and with the dimension of a 10-nm particle, which should have about 300-kDa mass assuming a typical value for protein density is closest to a tetramer (4 × 81 kDa). Previous studies show that some disulfide bonds are important for the functionality of prestin , although they are not required for dimerization as observed by gel electrophoresis .
Finally, the cell-free preparation formed from OHCs that we described would be useful in further studying the 10-nm particles and oligomerization of prestin. The uniform density and alignment of the 10-nm particles appear to make the cell-free preparation of OHCs logical for examining the effect of factors that control oligomerization, particularly that of cholesterol. We plan to address this issue by improving our preparation method.
Our study shows that AFM is a useful method in studying the structure of the plasma membrane when combined with cell-free preparations. We showed that the cytosolic face of the plasma membrane of the lateral wall of OHCs has particles that match the 10-nm particles reported by EM studies. We further provided evidence that these particles have immunoreactivity to prestin antibodies. These observations indicate that the 10-nm particles are a focus of further studies to clarify structure–function relationship of electromotility. The combination of a cell-free preparation and AFM could be useful methods for such studies. We may need to improve the cell-free preparation to minimize disturbance to membrane proteins to address the issue regarding packing density and pattern of these particles as well as to make the method viable for functional studies.
We thank Dr. Bechara Kachar for kindly providing us with prestin antibodies. We also thank Drs. Ronald Petralia and Paul Gallant for useful comments. This work is supported by the intramural research program of NIDCD, NIH.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.