Prestin in HEK cells is an obligate tetramer

Abstract

The unusual membrane motor protein prestin is essential for mammalian hearing and for the survival of cochlear outer hair cells. While prestin has been demonstrated to be a homooligomer, by Western blot and FRET analyses, the stoichiometry of self association is unclear. Prestin, coupled to the enhanced green fluorescent protein, was synthesized and membrane targeted in human embryonic kidney cells by plasmid transfection. Fragments of membrane containing immobilized fluorescent molecules were isolated by osmotic lysis. Diffraction-limited fluorescent spots consistent in size with single molecules were observed. Under continuous excitation, the spots bleached to background in sequential and approximately equal-amplitude steps. The average step count to background levels was 2.7. A binomial model of prestin oligomerization indicated that prestin was most likely a tetramer, and that a fraction of the green fluorescent protein molecules was dark. As a positive control, the same procedure was applied to cells transfected with plasmids coding for the human cyclic nucleotide-gated ion channel A3 subunit (again coupled to the enhanced green fluorescent protein), which is an obligate tetramer. The average step count for this molecule was also 2.7. This result implies that in cell membranes prestin oligomerizes to a tetramer.

prestin is a 744-amino acid membrane protein found in the lateral wall plasma membrane of cochlea outer hair cells (OHCs), and is a member of the Slc26a family of anion transporters. Its importance to mammalian hearing, and OHC survival, has been revealed by knockout and knockin studies in mice (Cheatham et al. 2004; Dallos et al. 2008; Liberman et al. 2002), although its role in hearing is far from clear.

Evidence for the oligomerization of prestin comes from several sources. Freeze fracture images of the OHC lateral wall plasma membrane revealed a dense, regularly spaced array of 10- to 12-nm round particles (He et al. 2010; Kalinec et al. 1992; Koppl et al. 2004). These particles, plausibly located in the OHC lateral wall plasma membrane, are thought to be prestin. However, the particles are 4–5 times the expected size of a 744-amino acid membrane protein. A Western blot analysis of prestin purified from expression systems suggested dimer and tetramer states, with the tetramer state being more easily dissociated (Zheng et al. 2006). However, another study suggested that dimers were the normal configuration of prestin and some other Slc26 family members (Detro-Dassen et al. 2008). Förster resonance energy transfer (FRET) analyses of prestin coupled to fluorescent proteins also suggested the existence of homooligomeric interaction, but did not provide any information on stoichiometry (Currall et al. in press; Gleitsman et al. 2009; Greeson et al. 2006; Navaratnam et al. 2005; Wu et al. 2007). A low-resolution electron density map of purified prestin exhibited fourfold symmetry, consistent with a tetramer (Mio et al. 2008). Overall, the experimental results have been equivocal on the question of stoichiometry. Alternate approaches to the question are clearly needed.

Our approach takes advantage of recent advances in the detection and analysis of single-molecule fluorescence. Detection of single-molecule fluorescence requires very high sensitivity detection methods with low background noise, such as photomultipliers or cooled electron-multiplying charge-coupled device cameras. Single-molecule detection, imaging, and tracking have been applied to numerous questions, including the mobility of neurotransmitter and odorant receptors (Heine et al. 2008; Jacquier et al. 2006), protein conformation change (Harms et al. 2003), and protein turnover (Leake et al. 2006). A more relevant application is the application of single-molecule sequential bleaching to the stoichiometry of ion channel subunit oligomerization (Harms et al. 2001; Ji et al. 2008; Ulbrich and Isacoff 2007). Briefly, single fluorescently coupled molecules are isolated and exposed to continuous excitation. The fluorescence is proportional to the number of active fluorophores in the molecule, i.e., the number of subunits. If the number is small, discrete equal-amplitude step decreases in fluorescence (bleaching) are observed until all fluorophores are bleached and the fluorescence is indistinguishable from background levels. The step count corresponds to the number of monomers in the molecule. The technique has been applied to L-type calcium channels (Harms et al. 2001), cyclic nucleotide-gated sodium channels (Ulbrich and Isacoff 2007), calcium release-activated calcium channels (Ji et al. 2008), and a voltage-gated proton channel (Tombola et al. 2008).

In this study, we applied the sequential bleaching method to the problem of prestin stoichiometry. We first established that we could measure single-molecule fluorescence by imaging a preparation of known single fluorescent molecules. We used streptavidin-Alexa Fluor 488 bound to biotin-coated coverslips as our test preparation.

We then used a preparation of membranes of human embryonic kidney cells (HEK cells) containing prestin coupled C-terminal to the enhanced green fluorescent protein (eGFP). We observed that the average step count was close to 2.7. As a positive control, we performed identical measurements on HEK cells synthesizing the A3 subunit of the human cyclic nucleotide-gated sodium channel (CNGA3, from cone photoreceptors), again coupled C-terminal to eGFP. CNG proteins form tetrameric ion channels in rod and cone photoreceptors composed of A and B subunits (Liu et al. 1996). The B subunits appear to be modulatory, while the A subunits form functional tetrameric ion channels in the absence of B subunits (Bonigk et al. 1993; Kaupp et al. 1989). We found the same average step count for this molecule. We therefore conclude that prestin in membranes is also an obligate tetramer.

MATERIALS AND METHODS

Observation of Alexa Fluor 488 fluorescence.

Polyethylene glycol-coated glass coverslips with a low density of covalently attached biotin were obtained from MicroSurfaces (Austin, TX). Coverslips were coated with 0.1 μg/ml streptavidin-Alexa Fluor 488 (Invitrogen, Carlsbad, CA) in high-purity water for 2–5 s, rinsed in high-purity water exposure, drained, and mounted for observation on an Olympus IX-70 inverted fluorescence microscope via a 100 magnification, 1.40 numerical aperture objective. Excitation and observation were performed using a mercury excitation source filtered by a standard fluorescein filter set. Areas for observation were first bleached to near darkness before acquiring images to minimize the number of fluorescent spots. Each streptavidin molecule had between two and four covalently linked Alexa Fluor 488 molecules, according to the manufacturer.

Sequences of fluorescence images were acquired using an Andor Technology (Belfast, N. Ireland) DU-897E cooled, back-thinned, electron-multiplying charged-coupled device camera (quantum efficiency >0.9). Images were acquired typically for 2,000 frames at 0.18 s/frame over a 128 by 128 pixel image field equivalent to 19.2 by 19.2 μm, with the camera electronics cooled to −60°C and the electron-multiplication gain set to between 50 and 80. Images were acquired in a darkened room with the microscope shrouded by a darkroom cloth. Measurements were made at room temperature.

Analysis of image sequences was performed using a procedure developed in MatLAB (The MathWorks, Natick, MA). Acquired images were smoothed using two passes of a three-point low-pass spatial filter. A 5 pixel by 5 pixel region of interest (ROI) (750 nm by 750 nm) was selected to encompass putative single-molecule fluorescent points. The summed fluorescence (in arbitrary units) was calculated for each image in a sequence using the ROI as a mask. The corner points (0, 0; 0, 5; 5, 0; 5, 5) were set to zero because they contribute little to the signal from a single fluorescent molecule, which appeared to have a radius of ∼700 nm. The summed ROI fluorescence in each frame was then divided by 21 to obtain the average fluorescence, and the values were plotted and stored in temporal order.

Step decreases in fluorescence were selected only if the decreases occurred over one or at most two frames. Step amplitudes were measured using the average fluorescence values before and after the step. The step amplitudes were accumulated in histogram form, and the histogram was fitted to a Gaussian distribution:

$$y=\frac{A}{w\sqrt{2\pi }}\exp \left(-\frac{{\left(x-xc\right)}^2}{2w^2}\right)$$ (1)

where xc is the average step amplitude (in counts), w is the standard deviation, and A is a positive constant.

In some experiments, 0.1 μg/ml streptavidin-Cy2 (Invitrogen) was used in place of the Alexa Fluor 488 conjugate.

Plasmids and transfection.

A plasmid encoding gerbil prestin in frame with eGFP was a gift from Peter Dallos (Northwestern Univ.). A plasmid encoding human CNGA3 (isoform 1), coupled C-terminal to eGFP, was a gift from Jacqueline Tanaka (Univ. of Pennsylvania). Cells (HEK-293) were cultured on glass-bottomed cultured dishes (MatTek, Ashland, MA) precoated with rat-tail collagen (BD Biosciences, Bedford, MA). Transfections of 60–80% confluent cells were performed using Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. Lipofectamine was removed after 6–8 h.

Osmotic lysis method.

Dishes of cells were observed 12–24 h after transfection. Cells were lysed using the method of Ziegler et al. (1998), as adapted by Murakoshi et al. (2006), with modifications as follows. Cells were exposed to a hypoosmotic buffer in the cold for 30 min. The buffer contained 4 mM PIPES and 30 mM KCl (pH 6.2, 80 mOsm). Cells were then subjected repeatedly to a stream of the same buffer delivered via a blunted 28-gauge hypodermic needle. Dishes were allowed to come to room temperature before observations were made. No intact cells could be observed on the culture dish after this treatment.

Observation of eGFP fluorescence.

Presumed membrane fragments containing fluorescence were observed on the stage of an Olympus IX-70 inverted fluorescence microscope. Excitation and imaging were as before except that during acquisition a neutral density filter attenuated the excitation to retard bleaching. Images were acquired typically for 800 frames at 0.2 s/frame over a 128 by 128 pixel image field, with the camera electronics cooled to −70°C and the electron-multiplication gain set to between 80 and 200.

Analysis of single-molecule eGFP fluorescence.

ROIs containing putative single molecule fluorescence were distinguished from random noise by following the ROI over sequences of images. Consecutive pairs of images were averaged and the resulting images spatially filtered as before. Background subtraction was performed by estimating the background fluorescence of each image as a function of time, using a 9 pixel by 9 pixel ROI placed away from potential single molecules. Putative single-molecule fluorescent points were analyzed as described earlier. A fraction (0.8) of the background fluorescence was subtracted from the ROI fluorescence by time function, so that the values remained positive.

On the first analysis pass, the step decrease magnitudes were determined as before and were plotted in histogram form. The distribution of step sizes was fitted to the sum of two Gaussian functions as follows:

$$y=\frac{A}{w\sqrt{2\pi }}\exp \left(-\frac{{\left(x-xc\right)}^2}{2w^2}\right)+\frac{A_2}{2w\sqrt{2\pi }}\exp \left(-\frac{{\left(x-2xc\right)}^2}{8w^2}\right)$$ (2)

where A₁ and A₂ are positive constants, and xc and w have the same meanings as before.

For each ROI, the step count was determined by inspection, informed by the average step size xc as determined above. The fluorescence in most ROIs reached background fluorescence levels, as determined by sampling the fluorescence of ROIs adjacent to the point, within the 800 frames. Only records in which background levels were reached were analyzed. All points meeting the above criteria were included in the analysis. The distribution of step counts was plotted in histogram form. The resulting histogram was fitted to a binomial sampling distribution as follows:

$$y=A\left(\frac{n!}{x!\left(n-x\right)!}\right)p^x{q}^{\left(n-x\right)}$$ (3)

where p is a probability, q = 1-p, A is a positive constant, and n is the integer sample size. The average of the binomial sampling distribution, μ, is np. For a sample size of n, the probability p was then calculated as μ/n.

RESULTS

Size determination of diffraction limited fluorescence points.

To assure ourselves that we were able to detect single fluorescent molecules, we used streptavidin-Alexa Fluor 488 applied to biotin-coated slides, as described in materials and methods. Extensive (30–90 s) prebleaching of the slide reduced the fluorescence to multiple discrete points, as illustrated in Fig. 1A. Discrete fluorescent points were readily identifiable, as seen in Fig. 1B. The time course of average fluorescence in the ROI was determined frame by frame using 5 × 5 pixel ROIs, as shown in Fig. 1B (750 nm at the specimen), and plotted as a function of time. Examples of the time course of fluorescence in ROIs are shown in Fig. 2A. Figure 2A1 shows the most common observation, of a discrete step decrease in fluorescence that occurred over 1 or 2 frames. Fig. 2A2 shows an example of an occasional observation, of two consecutive, approximately equal-amplitude step decreases in fluorescence. Figure 2A3 shows an example of the observation of repetitive on-off fluorescence steps, of equal amplitude. This phenomenon, called blinking, is characteristic of small-molecule fluorophores such as the Alexa Fluor dyes (Aitken et al. 2008).

Fig. 1.

Determination of the diffraction-limited single-molecule fluorescence point size. A: representative image field of a streptavidin-Alexa Fluor 488-labeled biotin-coated coverslip after extensive prebleaching. Imaging conditions were as set out in materials and methods. The image of 128 × 128 pixels represents 19.2 × 19.2 μm² at the coverslip. B: higher magnification field of the same image showing representative placement of regions of interest (ROIs) (5 × 5 pixel, 750 × 750 nm) demarcating putative points of single-molecule fluorescence.

Fig. 2.

Evidence for imaging of single fluorescent molecules. The red lines are the segments that were averaged to estimate the step size. A: representative time courses of fluorescence from the same experiment as Fig. 1. The scale bars represent 5 s (time) and 100 counts (amplitude). A1: typical step decrease in fluorescence. A2: occasional observation of 2 approximately equal step decreases. A3: blinking by a single molecule. B: histogram of 95 observed fluorescence step decreases in a single experiment. The fitted Gaussian distribution parameters are xc = 87.49, w = 37.48, A = 467.13 (R² = 0.86). C: histogram of 339 observed fluorescence decrease steps in a single experiment using streptavidin-Cy2 instead of streptavidin-Alexa Fluor 488 (3 coverslips). The fitted Gaussian distribution parameters are xc = 205.99, w = 71.53, A = 3345.77 (R² = 0.96).

The amplitudes of 95 step fluorescence decreases from a single experiment are plotted in histogram form in Fig. 2B. The histogram was well fit by a Gaussian distribution (Eq. 1, R² = 0.86). The fit implies that the step decreases were unitary events, subject to measurement error, which is consistent with the assumption that the steps are the bleaching events of single fluorescent molecules. We experimented with larger (7 × 7 and 9 × 9 pixel) ROIs, but the signal-to-noise ratio of the records was lower.

We repeated the experiment using streptavidin-Cy2 instead of streptavidin-Alexa Fluor 488. The histogram of 339 step amplitudes from one such experiment is shown in Fig. 2C. Again, the histogram was well fit by a Gaussian distribution (R² = 0.96).

Sequential bleaching of prestin.

We first examined the fluorescence of HEK cells synthesizing prestin-eGFP. We observed points of fluorescence consistent in size with single molecules, as determined above, as well as larger points that may have been macromolecular assemblies. However, many of the points were mobile, and some appeared to undergo endocytosis, which made analysis of bleaching events unreliable.

We therefore removed the cells by osmotic swelling and lysis, as described in materials and methods. Inspection of the culture dish under transmitted light revealed no intact, attached cells. However, UV excitation revealed points of fluorescence, again consistent in size with single molecules, which were not mobile (Fig. 3), as well as larger points as before.

Fig. 3.

Punctate fluorescent points observed in HEK cell membranes after hypoosmotic lysis, showing representative ROIs demarcating putative points of single-molecule fluorescence. The image of 128 × 128 pixels represents 19.2 × 19.2 μm² at the culture dish.

Image sequences of the fluorescent points were acquired as described in materials and methods, with background subtraction. ROIs containing putative single molecules were analyzed as follows and as depicted in Fig. 4, for three experiments. We analyzed only points for which the fluorescence decreased to background levels, estimated as described in materials and methods. Figure 4A is an example of a typical point that decreased to the background fluorescence level in four discernable steps. We also observed numerous points for which the fluorescence decreased in three steps (Fig. 4B1, 4B2). We also observed some points with two steps (Fig. 4C) and some with only one step (Fig. 4D).

Fig. 4.

Representative time courses of fluorescence of single-molecule ROIs in the isolated membrane of a prestin-transfected cell. A: example of 4 steps. B1 and B2: examples of 3 steps. Arrow in record B1 indicates on-blink. Arrow in record B2 indicates transient dwell at an intermediate level. C: example of 2 steps. Arrow indicates transient off-blink. D: example of single step. In all records, the red lines represent the average fluorescence in that time range used to measure the step amplitude. The scale bars represent 10 s (time) and 100 average counts per pixel (amplitude).

Step amplitudes were measured as described and accumulated in histogram form (Fig. 5A1, B1, C1). Histograms were best fit by the double Gaussian distribution (Eq. 2) with only a small fraction of the steps of amplitude 2xc (i.e., A₂<<A₁). We interpreted this as evidence that we were observing mainly a single population of bleaching events, with the occasional near-simultaneous pairing of events.

Fig. 5.

Histograms of average step size (A1, B1, C1) and step counts (A2, B2, C2) from 3 experiments. The fit parameters are shown.

We then determined the step count for each point, using the single-step amplitude xc derived from the amplitude histogram to resolve any ambiguities. The step counts were accumulated and displayed in histogram form (Fig. 5A2, B2, C2). The average step counts in the three experiments were 2.69, 2.71, and 2.71. The step count histogram for each experiment was then fit to a binomial distribution (Eq. 3). As shown in Fig. 5, the fits were remarkably similar, with sample size n = 4, and the probability p between 0.65 and 0.68. This occurred even though the electron-multiplying gains and the attenuation of the excitation source were different for the three experiments.

Sequential bleaching of CNGA3.

For comparison, we transfected cells with a plasmid expressing a C-terminal construct of the human CNGA3 ion channel coupled C-terminal to eGFP, a known tetramer. We performed the same experiment and analyses on these cells. The results (Fig. 6) were almost identical to those observed for prestin-eGFP. The average step counts were 2.59 and 2.74, and the binomial distribution fits were n = 4, p = 0.65 and 0.68.

Fig. 6.

Histograms of average step size (A1, B1) and step counts (A2, B2) from 2 experiments using CNGA3-eGFP. The fit parameters are shown.

DISCUSSION

The binomial model predicts that prestin is a tetramer.

The success of the binomial model in fitting the step count distribution of this experiment suggests a simple model to explain the results. In the binomial sampling model, small samples of size n are withdrawn from a large pool. Here, each molecule represents a sample and n is the stoichiometry. In the binomial model, not all units are good, only a fraction p. The remainder, q = 1-p, are defective. The distribution of observation of good items in repeated sampling is given by Eq. 3. In our experiment, good is interpreted as able to fluoresce. The optimum sample size n to fit the data was four, which implies that the molecules, prestin and CNGA3, are tetramers. Since CNGA3 is a known tetramer, our model predicts that prestin is also a tetramer.

The value of p in this model, between 0.65 and 0.68, implies that only 65–68% of eGFP molecules were able to fluoresce, the remainder being dark. This explains why we observed a range of step counts in our sample, from four to one, which is consistent with a binomial sampling process with p < 1. As is clear from Figs. 5 and 6, we occasionally observed points with step counts greater than four. We interpret these points as having contained more than one prestin molecule, and therefore did not include them in the analysis.

Alternative explanations of the results.

We considered several other models that might fit the results. Dimer and trimer configurations were rejected because of the high frequency of points in which the step count was greater than three. Random association of monomers into oligomers would produce a uniform distribution of step counts, unlike our observations. As mentioned in the previous section, we observed a small number of points with greater than four steps. We examined the possibility that prestin is a pentamer (i.e., n = 5) and included them in a re-analysis. However, the analysis suggested that p is ∼0.5, i.e., that about half of the eGFP molecules were dark, which we considered to be unlikely, especially since p was 0.65–0.7 for the positive control CNGA3, for which the average step count was identical to that of prestin.

Advantages of this approach.

The advantages of this approach are numerous. Prestin molecules are in the lipid plasma membrane and are therefore closer to their native state, unlike Western blots or electron density mapping, in which the molecules are in aqueous media (Westerns) or none at all (electron density mapping). The molecules are also isolated from each other, unlike FRET analyses, which required high densities for adequate signal levels.

The HEK membrane is not necessarily identical to OHC lateral wall membrane, which has unusually large cholesterol levels (de Monvel et al. 2006; Rajagopalan et al. 2007). However, the electrical properties of prestin in HEK cells, even when coupled to eGFP, are essentially the same as in the OHC (Deak et al. 2005; Santos-Sacchi and Navarrete 2002), thus the configuration of prestin in HEK cell membranes is likely to be representative of its natural state.

Sources of dark eGFPs.

Our binomial model predicts that the molecules are tetramers, based on the optimum fit with n equal to 4. However, it also predicts a larger-than-expected proportion of dark eGFP molecules (1-p). Previous work has suggested that at least 20% of eGFPs are dark (Ulbrich and Isacoff 2007). Our method requires some focusing on the fluorescent points, because the preparation is not level over the dish and possibly the membranes are themselves uneven, which was not the case with the Ulbrich and Isacoff experiment. The excitation intensities used in our experiments were greatly attenuated to delay bleaching events and thereby facilitate the analysis. However, the excitation intensities could not be reduced further, or the bleaching events themselves became undetectable. It is likely that some bleaching of eGFP occurred before recording could begin, which may account for some part of the 30–35% dark eGFP molecules observed.

We confirmed this speculation by re-analyzing one of the experiments, the one depicted in Fig. 5C. We compared the results of the analysis, assuming n = 4, with the same analysis but starting at 10 s, 20 s, or 50 s after the acquisition had begun. This procedure simulated the effects of prebleaching the fluorophores. Our model predicts that the probability p should decrease with prebleach duration, but the average step size should be unaffected. As shown in Fig. 7, this was indeed the case. The average step size remained remarkably constant, while the p value decreased with a single exponential time course.

Fig. 7.

Time course of average step size and binomial parameter p in a simulated prebleaching experiment. The details are in discussion. Fit parameters are as follows: average step size, slope = −0.72, R = −0.89; p, τ = 18.49 ± 4.73, R² = 0.99.

Prestin mechanisms.

The results of this study do not directly address the mechanism of prestin conformation change. Nor do they help decide the question of cooperativity between prestin subunits. Wang et al. (2010) argued on numerical grounds that each subunit operates independently. However, Detro-Dassen et al. (2008) found that prestin molecules formed in cells coexpressing two prestin mutations with different electrical properties had intermediate electrical properties in the aggregate. This, they suggested, indicates that one subunit can influence another, which implies cooperativity. It is likely to require single-molecule methods, such as the one demonstrated here, to resolve this important structure-function question.

GRANTS

This work was supported by a grant from the State of Nebraska LB 692 fund to R. Hallworth and National Institutes of Health (NIH) Grant GM-085776 to M. G. Nichols. Research was conducted in a facility constructed with support from Research Facilities Improvement Program C06 RR-17417-01 from the National Center for Research Resources of the NIH.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: R.H. and M.G.N. conception and design of research; R.H. performed experiments; R.H. analyzed data; R.H. and M.G.N. interpreted results of experiments; R.H. prepared figures; R.H. drafted manuscript; R.H. and M.G.N. edited and revised manuscript; R.H. and M.G.N. approved final version of manuscript.