Neuron, Vol. 7, 985-994, December, 1991, Copyright 0 1991 by Cell Press Tip-link Integrity and Mechanical Transduction in Vertebrate Hair Cells

John A. Assad,*+ Gordon M. G. Shepherd,* This suggests that the gating springs are not rigid ele- and David P. Corey*511 ments, but can be slack-that they can pull but not *Department of Neurobiology push on the channels (Corey and Hudspeth, 1983). *Program in Neuroscience The structural correlate of this process has not been Harvard Medical School well established. A simple model has evolved from Boston, Massachusetts 02115 several independent observations. First, measure- SDepartment of Neurology ment of current flow near moving bundles indicated Massachusetts General Hospital thatthetransductionchannelsareatornearthetipsof Boston, Massachusetts 02114 the (Hudspeth, 1982). While this has been IINeuroscience Group challenged by measurements with a Ca*+ indicator Howard Hughes Medical Institute dye (Ohmori, 1988), two additional experirnents have corroborated the localization of the channels at the tips (Huang and Corey, 1990, Biophys. Sot., abstract; Summary Jaramillo and Hudspeth, 1991). Second, the discovery of fine filaments between the tips of adjacent ster- An attractive hypothesis for hair-cell transduction is that eocilia led to the suggestion that these “tip links”were fine, filamentous “tip links” pull directly on mechani- the actual mechanical linkages to the channels (Pick- cally sensitive ion channels located at the tips of the les et al., 1984). The geometry of the bundle is such stereocilia. We tested the involvement of tip links in the that excitatory displacements would stretch the tip transduction process by treating bundles with a BAPTA- links and apply tension to the channels; inhibitory buffered, low-Ca*+ saline (1O-v M). BAPTA abolished the displacements would relax them. All vertebrate hair transduction current in a few hundred milliseconds. cells so far examined-from different species and dif- BAPTA treatment for a few seconds eliminated the tip ferent organs whose stereociliary morphology may links observed by either scanning or transmission elec- otherwise vary-possess tip links. The tip-lr,nk hypoth- tron microscopy. BAPTA also eliminated the voltage- esis for transduction is thus extremely attractive, yet dependent movement and caused a positive bundle dis- direct evidence for it is limited. placement of 133 nm, in quantitative agreement with a In this paper we directly implicate the tip links in model for regulation of tension. We conclude that tip the transduction process, with the finding that a brief links convey tension to the transduction channels of hair treatment of low extracellular Ca2+ destroys both the cells. tip links and the mechanical sensitivity. In dissociated cells, moreover, low Ca*+ both abolishes the voltage- Introduction dependent bundle movement driven by the cells’ active regulation of gating spring tension and causes When the mechanosensitive bundle of a vertebrate the bundle to relax forward by 133 nm, 1ir-rquantita- is displaced in the positive direction (toward tive agreement with the idea that low Ca2+ destroys the taller stereocilia), transduction channels open and transduction by cutting the attachments to the ion allow the flow of positive ions into the cell. The open- channels. ing of a channel is thought to result from an increase of mechanical tension on the channel protein itself. Results The principal evidence for direct mechanical gating of these channels is that the transduction process is Abolition of Transduction by low Ca*+ extremely rapid (Corey and Hudspeth, 1979a), that the It has long been recognized that Ca*+ in the solution opening and closing rates depend on the size of the bathingthe hair bundles is required for hair-cell trans- stimulus (Corey and Hudspeth, 1983; Crawford et al., duction (Sand, 1975; Corey and Hudspeth, 1979b; 1989), and that the mechanical complianceof a bundle Crawford et al., 1991). While earlier experiments sug- includes a component that matches the opening of gested that Ca*+ might carry the receptor current the channels (Howard and Hudspeth, 1988). In this (Sand, 1975), more recent work views Ca2+ as a neces- view, the open probability is a direct function of ten- sary cofactor for the transduction apparatus (Craw- sion, conveyed by an elastic “gating spring.” A pecu- ford et al., 1991). We have reexamined the Ca2+ depen- liarity of the gating kinetics is that channel opening is dence with whole-cell voltage clamp, direct bundle progressively speeded by larger positive displace- stimulation, and rapid application of tes,t solutions. ments, whereas the closing rate is independent of the Figure 1 shows transduction currents of single cells in stimulus for sufficiently large negative displacements. response to a triangle-wave stimulus of 1.0 pm peak- to-peak amplitude; this was a saturating stimulus for these cells. The bath contained a normal frog saline + Present address: Department of Physiology and Center for Vi- sual Science, University of Rochester, Rochester, New York with 4 mM Ca*+. Halfway through the record, a low- 14642. Ca*+ saline (lOmg M; buffered with 5 mM BAPTA) was Neur0n 986

I I I 1 Time (set)

Figure 1. Abolition of Transduction by Low Ca’+ (A-C) Transduction current in 3 different cells elicited by a 1 pm peak-to-peak, tnangk-wave displacement ot the bundle (bottom trace) A 5 mM BAPTA solution was delivered by a pressure pipette positioned about 20 @m from the cell. at the time Indicated by the bar In all 3 cases, transduction current was abolished by the BAPTA treatment and did not return for the duration ot the recording, usually several minutes after the exposure to low Ca2+. Membrane potential was maintained at -80 mV by the patch clamp. The bath contained the external recording solution, with 4 mM Caz+.

puffed onto the cell from a distance of about 20 pm. curve-which relates receptor current to displace- The bundlewas held firmly bythe stimulus probeand ments-toward more negative positions, so that more continued to be driven by the triangle-wave stimulus. channels are open at rest (Assad et al., 1989; Crawford As the wave of low Ca2+ reached the bundle, the trans- et al., 1989, 1991). In extreme cases, it is possible that duction currentwas initially increased by about lOO%, such a shift is so large that all the channels are open a consequence of relieving a voltage-dependent block with zerodisplacement and that a cell appears insensi- of the channels by Caz+ (Assad and Corey, unpub- tive to mechanical stimuli, while in fact a sufficiently lished data; Crawford et al., 1991). In as little as 50 ms, large negative displacement could close channels. however, the transduction currentwas abolished, and That does not appear to account for the loss of sensi- it was not restored even several minutes after the tivity in this experiment, because there was not evi- BAPTA solution diffused away. In Figures IA and IB dence of a progressive shift of the I(X) curve, because the total membrane current increased, at least tran- the stimulus included large negative excursions that siently, during the exposure to low Ca*+. This increase were ineffectual in closing channels, and because the is due, at least in part, to an observed shift in the effect was not reversible upon return to normal Ca*- activation of basolaterally situated voltage-dependent concentration. It seems more likely that some part of Ca2+ channels (Roberts et al., 1990) to more negative the transduction apparatus was actually broken by potentials- perhaps as a result of alleviation of screen- lowering the Ca*+ concentration. ing charge-coupled with an increased monovalent ion flux through these channels (Assad and Corey, Abolition of Tip links by low CaZ+ unpublished data; Hess et al., 1986). To determinewhat part of the transduction apparatus We and others have found that lowering the Ca2+ was broken, hair cells were similarly treated with low concentration inside the stereocilia shifts the I(X) Ca*+ and prepared for scanning electron microscopy Hair Cell Transduction 987

Control

Figure 2. Scanning Electron Micrographs of Stereocilia Treated with CaZ+ or BAPTA Ca2+-treated sacculi (left) were dissected in a 0.1 mM Caz+ solution and then placed in 4 mM Ca*+ before fixation in 4 mM Caz+. BAPTA-treated sacculi (right) were dissected in the same way, but placed in a 5 mM BAPTA solution for 10 s before fixation in 4 mM Caz+. Bar, 500 nm. (SEM). Sacculi were dissected as for physiological ex- periments and treated to remove their otolithic mem- branes, but were not dissociated. Experimental sac- culi were transferred to a solution buffered with 5 mM BAPTA for about 10 s and then immediately returned to normal saline. Control sacculi were transferred to normal saline in the same manner. Both sets were fixed with glutaraldehyde and 0~0~ and processed for field emission SEM. In order to photograph a representative sample of both control and experimental bundles, maculae were viewed at low magnification (3000x) and bun- dles were chosen more or less randomly from all re- gions. At the low magnification it was not possible to observe tip links; thus, their presence could not bias the choice of a cell. Bundles were photographed at Figure 3. Intact Tip Links following Detergent Treatment high magnification (50,000x), from an angle approxi- Specimen was dissected in 0.1 mM CaI+ and then maintained in control saline containing 4 mM Ca2+ at all times prior to fixation. mately perpendicular to the bevel of the tips. Twelve Triton X-100 (2%) was added to the fixative for the final 20 min to20 bundles were photographed from each of 4 mac- of fixation. Bar, 500 nm. ulae. Figure 2 shows representative images of control (left) and BAPTA-treated (right) bundles taken with an accelerating voltage of 4 kV. The stereocilia in both ular, the bundles were not splayed or otherwise dis- samples appearwell preserved, with no bending. The rupted by the BAPTA treatment. lumpy or convoluted appearance of the stereocilia The presence of tip links was more quantitatively may represent condensed glycocalyx or membrane assessed by measuring the proportion of stereocilia wrinkled bydifferential shrinkage,although transmis- bearing tip links in each image. Sixty-three photo- sion electron microscopy (TEM; see below) supports graphic images were coded, randomized, and judged the former interpretation. At higher accelerating volt- by four observers. The proportion of tip links was ages (35 kV), these same stereocilia appeared smooth calculatedastheaveragenumberoftiplinkscounted, and slightly translucent and tip links were less dis- divided by the number of tips for which links could tinct. have been seen. Altogether, 420 tip links were In control bundles, tip links can be observed ex- counted, out of more than 1800 stereociliary pairs. tending from the tips of most stereocilia to the sides The results are presented as a histogram in Figure 4, of the tallest adjacent stereocilia. Links are arranged showing the percentage of tip links per bundle. In along columns of stereocilia, parallel to the physiolog- controls, tip links were generally abundant; the aver- ical axis of the bundle, as previously described (Pick- age was 39% rf- 25% (SD, N = 38). In BAPTA-treated les et al., 1984, 1989). The tip links viewed by SEM bundles, tip links were rarely observed (1.4% k 1.9%, appear thicker and shorterthan those viewed by TEM N = 25). Moreover, those that were observed were (see below), possibly because of the gold-palladium questionable: in no case did all four observers agree coating or the condensation of the glycocalyx onto that a BAPTA-treated bundle contained a tip link. the membrane. However, the difference raises the Analysis of similarly treated maculae by TEM concern that the links observed in SEM are not bona yielded similar results. Figures 5a-5d show represen- fide tip links, but perhaps tubes of membrane drawn tative tip links between stereociliary pairs from con- fromthetipsof stereocilia.Toexcludethispossibility, trol bundles. The links characteristically inserted into several sacculi were treated with a detergent (Triton osmophilic densities at either end. Representative X-100,2%) to remove cell membranes after the glutar- BAPTA-treated stereocilia are shown in Figures 5f-5j. aldehyde fixation (Figure 3). Triton-treated stereocilia The proportion of tip links was scored by photo- differed from untreated bundles in that their surfaces graphing every stereociliary pair for which a tip link were smooth, perhaps representing the underlying might have been observed, as judged by the presence actin cores. Yet the tip links of treated bundles were, of both upper and lower osmophilic densities. Be- if anything, more apparent. Thus the links in field cause the densities are thicker than the tip link and emission SEM images most likely represent the tip may occur in several sections, the actual proportion links observed with TEM and are not a lipid artifact. of tip links was probably underestimated. The scored It is immediately apparent from Figure 2 that the results are shown in Figure 5e. Tip links were almost main morphological consequence of BAPTA treat- never seen in BAPTA-treated specimens (1.3% * 1.8% ment was the loss of tip links. In each control bundle, SD; N = 58),whiletheyappeared in control specimens roughly half of the stereocilia have tip links, whereas in about half the stereociliary pairs where they might tip links are largely absent in BAPTA-treated bundles. have been seen (48% + 4% SD; N = 74). No other systematic difference was noticed; in partic- A common observation with TEM is membrane Hair Cell Transduction 989

Control “tenting’at the lower site of attachment of the tip link, where the membrane is seen rising away by approxi- mately 15 nm from the underlying osmophilic density that caps the actin core (Figure 5). This effect is pre- sumably due to tension in the tip link. The plroportion a, of stereocilia displaying tenting was similarly affected P 0 10 20 30 40 50 60 70 80 90 100 by treatment with BAPTA: tenting was observed in L 2 25 -, 61% + 5% (SD, N = 74) of control samples and 14%

5 k 10% (SD, N = 58) of BAPTA-treated samples. & 20 - In addition to tip links, stereocilia are connected by 5 mM BAPTA upper and lower lateral links, which span from each f 15- stereocilium to its 6 neighboring stereocilia (Bagger- 2 Sjobick and Wersall, 1973; Kimura, 1966). We ob- lo- served an effect of BAPTA treatment only on the tip 5- links. This was particularly evident when hair bundles were stained with 8% tannic acid to improve visualiza- 0- I , 1 I ( I , I I tion of the various links. In both control ancl test bun- 0 10 20 30 40 50 60 70 80 90 100 dles, the upper lateral links were readily identified, while the lower lateral links were barely evident (data % tip links not shown). The poor definition of the lower links may indicate their susceptibility to degradation in the Figure 4. Abolition of Tip Links by Low CaL+ enzyme treatment used to dissociate the otolithic The number of tip links in each micrograph was counted in a blind assay by four observers, and the average was calculated membrane from the kinocilia. This treatment, which for each micrograph. Data were expressed as a percentage of the leaves mechanical transduction intact, removes the number of stereociliary tips for which tip links could have been lower links (Jacobs and Hudspeth, 1990). seen. The histogram shows the distribution of this percentage for the control bundles (top) and the BAPTA-treated bundles (bottom). A total of 63 bundles were analyzed, 38 control and 25 Abolition of Active Bundle Movement by Low Ca2+ BAPTA-treated. Bin width, 5%. It thus seems that a principal consequence of BAPTA

b I d

igure 5. Transmission Electron Micrographs of Stereocilia Treated with Ca*+ or BAPTA :a2+-treated sacculi (a-d) were dissected in a 0.1 mM CaZ+ solution and then placed in 4 mM CaZ+ before fixation in 4 mM Ca2+. APTA-treated sacculi (f-j) were dissected in the same way, but placed in a 5 mM BAPTA solution for 3-4 s before fixation in 4 mM Ca’+. esults of scoring tip-link disposition, expressed as a proportion of the total number of stereociliary pairs forwhich both densities were pparent, are shown in (e). Error bars represent standard error of the mean. NeUrO” 990

Figure 6. Abolition of Acttve Bundle Move- ment by Low Ca>+ tn Three Different Cells Active bundle movement was elicited by 1 s depolarizations from -80 to +80 mV, repeated every 2 s, as indicated by the trace labeled V,,. Bundle movement was ana- lyzed using the automated procedure de- scribed in Experimental Procedures. The I I negative direction is shown as downward. 20 mM BAPTA At the time denoted by the bar in each ex- periment, a low-Cal’ solution, buffered with the indicated concentration of BAPTA, was gently puffed onto the cell from a pres surepipettepositionedatleast6Oumaway. In each case, the voltage-dependent move- mentwas irreversibly blocked within about 5 s after the initiation of the puff, and the bundle moved to a more positive position. The amplitude of the positive relaxation 4 mM BAPTA was measured 2 s after the initial move- ment of the bundle. In some cases the posi- tion then began to drift gradually in the negative direction, although the drift was not qualitatively different from the base- line drift normally observed in other active-bundle movement experiments (As- sad and Corey, 1992). The significance nt the large transient movements, occurring in either the positive (A) or the negative (B and C) direction at the initiation of the forward relaxation, was unclear. However, it is unlikely that they were due solely to 2 mM BAPTA mechanical artifacts, since their direction did not always correlate with the directton of flow of the puff.

set

treatment is the simultaneous loss of mechanical sen- dles by about 70 nm (Assad et al., 1989; Assad and sitivity and the loss of tip links, strongly suggesting Corey, 1992). If the gating springs were cut, in this that the tip links are an essential component of the view, the voltage-dependent movement would be transduction mechanism. The simplest interpretation abolished and the bundle would relax forward by is clearly that the tip links are the gating springs that about 130 nm to its rest position in the absence ot convey stress to the channels. Yet BAPTA could de- gating spring tension. Bundle position is thus a me- stroy mechanical sensitivity by disrupting some other chanical assay of gating spring integrity. element of the transduction apparatus, so it is im- An experiment designed to test these predictions portant to assess the integrity of the gating springs by was performed on voltage-clamped single cells, by an additional method. Tension in the gating springs gently puffing a BAPTA solution on them while the is manifested electrophysiologically, by the opening membrane potential alternated between +80 and -80 of ion channels; the first experiment showed that mV (Figure 6). The puffer pipettewas positioned about BAPTA abolished the transduction current. It is also 60 pm from the hair bundle to minimize turbulence. apparent mechanically, in that altered tension on In Figure 6A it was oriented with the stream directed channels can move a free-standing bundle by a frac- at the short stereocilia. In Figures 6B and 6C the stream tion of a micrometer (Assad et al., 1989). Our current was directed at the so that the puff would model of adaptation supposes that an intracellular tend to push the bundle in the negative direction. The “motor” element maintains constant resting tension movementofthetipofthe bundlewasmeasuredfrom in the gating springs (Howard and Hudspeth, 1987; the video record using the automated routine de- Hacohen et al., 1989; Assad and Corey, 1992). This scribed in Experimental Procedures. tension exertsaslight negative pull on the hair bundle In each case, the effect of BAPTA was to abolish the of about 130 nm; the pull can be increased bydepolar- voltage-dependent movement irreversibly. Regard- ization, to generate a movement of free-standing bun- less of the direction of the puff, the bundle then Hair Cell Transduction 991

moved in the positive direction to a new resting posi- The confirmation of tip links as part of the transduc- tion. The average movement, measured relative to the tion chain would obviously limit the possible place- rest position at -80 mV, was 133 + 70 nm (SD, N = 5). ment of the transduction channels. As in the original suggestion of Pickles et al. (1984), channels almost Discussion certainly occur near the point where the tip links con- tact the stereociliary membrane. As yet unclear is The experiments reported here support the notion whether channels are at the upper or lower end of that tip links are part of the mechanical chain that the tip link, or at both. The estimate of Howard and conveys tension to the transduction channels. Low- Hudspeth (1988), that there are roughly 85 channels Ca*+ saline (lOmg M; buffered with BAPTA) irreversibly per cell, is consistent with a channel at either one or eliminated the tip links, as observed with both scan- both ends of each tip link, given an average of 60 ning and transmission electron microscopy.The same stereocilia per bundle (Jacobs and Hudspeth, 1990). treatment irreversiblydestroyed thegatingsprings,or Forge et al. (1988) and Jacobs and Hudspeth (1990) some element mechanically in series with the gating have described particles in freeze fracture replicas at springs, as shown by the loss of mechanical sensitiv- thetipsof stereocilia. lfthese particles,which number ity, and the abolition of voltage-dependent bundle 4-6 per stereocilium or about 300 per cell, represent movement with concurrent positive displacement of channels, then it may be that not all of them are me- free-standing bundles. The average positive displace- chanicallyconnected. If the mechanical chain extends ment with BAPTA treatment closely matched that pre- from the actin core of one stereocilium to the actin dicted from aquantitative model for regulation of ten- core of the other, then there must be transmembrane sion in the gating springs. elements at both ends of a tip link. It is thus possible It remains possible that the elimination of tip links that the freeze-fracture particles represent integral and the loss of tension in the gating springs are coinci- membrane proteins that are not channels, but that dental-that both are effects of low Ca*+, but are other- help to tie the tip link to the cytoskeleton. wise unrelated. However, both were irreversible and The confirmation of tip links as part of the chain both occurred with exposures of a few seconds or would also constrain the possible placement of the less. motor element that underlies adaptation. A variety of While these experiments strongly suggest that tip physiological evidence has indicated that a cyto- links convey tension to the channels, it does not nec- plasmic motor maintains the resting tension in the essarily follow that they are the elastic elements that gating spring, perhaps by moving one end of the gat- were inferred from physiological measurements and ing spring (Howard and Hudspeth, 1987; Eatock et al., that are termed the gating springs. It may be that the 1987; Hacohen et al., 1989; Assad and Corey, 1992). If tip links are themselves relatively inextensible and the tip link is the gating spring, then one of its attach- that some other element in series is stretched with ments must move. Howard and Hudspeth (1987) have positive displacements. The tenting of stereociliary speculated that the upper attachment point can slip tips might represent the gating springs, for instance. to relax tension following positive displacements that However, the tip links are both longer and thinner stretch tip links, and can climb to restore tension after than other candidate structures and might be ex- negative displacements that relax the links. We might pected to stretch the most. It seems most plausible then expect to find the osmophilic density of the up- that the morphologically described tip links are the per tip-link insertion to be lowerwhen fixed after posi- physiologically defined gating springs. tive displacements and higher after negative displace- Several other groups have used electron micros- ments. We might also expect that relieving tension by copy to examine the vulnerability of tip links to vari- cutting tip links with BAPTAwould allow the densities ous treatments, although none has directly correlated to climb toward the tips of stereocilia. In preliminary the effects with the physiology of the hair cells. Acous- experiments, this seems to be the case (Shepherd et tic trauma apparently disrupts tip links, but onlywhen al., 1991, Sot. Cen. Physiol., abstract). the bundlesare nonspecificallydamaged aswell (Pick- What is the biochemical nature of the tip links? les et al., 1987). A variety of proteolytic enzymes have First, the persistence of tip links after solubilization no detectable effect on the tip links (Pickles et al., of the cell membrane with detergent argues that the 1990; Osborne and Comis, 1990). For instance, a prote- structures are not membranous and that they are ase was used in these experiments to loosen the oto- linked in some way to the underlying cytoskeleton. lithic membrane, but it did not disrupt mechanical We expect that the electron-dense plaques, situated transduction or tip links. A possible exception is elas- between the membrane and the actin core where the tase (Osborne and Comis, 1990). However, elastase links insert, represent this connection. Second, the causes splaying of the bundle as well (Osborne and extremely rapid action of low Ca2+ on the tip links Comis, 1990), which might secondarily disrupt the tip suggests that they are directly sensitive to Ca*+. It is links. The brief BAPTA exposures used in our experi- possible that the tip link is a dimeric structure held ments caused no apparent change in the gross mor- together by Ca*+-dependent interactions between ap- phology of the bundle: stereocilia remained unbent posing components, analogous to other Ca*+-depen- and continued to touch at their tips. dent cell adhesion molecules, such as cadherins (Jes- NeLlrOn 992

sell, 1988). Alternatively, the tip links may be polymers Video Microscopy of smaller subunits, like actin or tubulin. Howard et Camera and Recording Hair cells were observed with a 63x objective and DIC optics on al. (1988) have pointed out that the 5 nm diameter of a Zeiss IM-35 inverted mlcroscope. An image of the field wa\ the tip links is similar to that of elastin filaments and projected at high magnification onto the faceplate of a Hama- that elastin can stretch to twice its length, as would be matsu Newviron video camera (C2400 series). so that the video required if tip links were the gating springs. The CaZ+ field subtended about 20 pm. The camera output was observed dependence of the tip links may provide a means for on an oscilloscope, and gain and offset were adjusted to just below saturation. The video signal was recorded on a standard their biochemical identification and characterization. VHS video cassette recorder and, for shorter segments, directly onto optical memory disk (Panasonic TQ-2026). Experimental Procedures Measurement of Active Bundle Movement The spontaneous movements of a hair bundle that tallow depo- Physiological Recording larizing voltage steps were quantified as described elsewhere Dissociation (Assad et al., 1989; Assad and Corey, 1992). Intensity profiles 01 Single hair cells were dissociated from the sacculi ot adult bull- up to 10 lines drawn across the bundle image, each about 50 frogs, as described elsewhere (Assad and Corey, 1992). Briefly, pixels long and 5 pixels wide, were measured with an ITI-151 a saline resembling (120 mM NaCI, 2 mM KCI, 0.1 mM image processor and recorded on optical memory disk for up to CaC12, 3 mM dextrose, 5 mM HEPES [pH 7.251) was buffered to several thousand frames, using a C-language program written bv a free Ca2+ concentration of IO-‘M with 1 mM EDTA and dripped P. L. Huang and N. Hacohen. These data files werethen analyzed onto sacculi that had been surgically exposed. While the EDTA with a QuickBASIC program on an 80486 computer to determine solution had access to the basolateral surfaces of cells, tight the frame-to-frame movement. For each line, intensity profile5 junctions of the intact labyrinth presumably prevented the solu- were averaged for several initial frames to create reference pro- tion from reaching the hair bundles. In addition, otoconia in files. Then, for each new frame, the reference profile was shifted each sacculus may have helped maintain a normal endolym- laterally to match each new profile, mlnimlzing the squared dit- phatic Ca2+ concentration near the bundles. After 15 min, sacculi ferences wtth Newton’s method. This method matched profiles were removed from the animal and further dissected, then within about 0.5 nm; the overall noise In the method, measured treated for 30 min with a protease solution (75 Kg/ml; type XXIV; with static images, was typically 4 nm. The shift needed to match SigmaChemical Co.) in salinecontaining0.1 mM CaCI,, to loosen profiles was then taken as the movement ot the bundle. the attachments to the otolithic membranes. Following removal ofthe otolithic membranes, cells were flicked out of the maculae into saline containing 1 mM CaCI>, 0.4 mM M&I,, and 40 pg/ml Scanning Electron Microscopy DNAse I (Worthington Biochemical Co.) and were allowed to Preparation and Fixation settle onto the clean glass bottom of the recording chamber. All Both sacculi were removrd tram the dn~mal and enzymatlcall\ steps of the dissociation were performed at room temperature. treated to tooben the otolithic membranes, as described abovr. Recording After the membranes had been peeled away, the sacculi were Membrane currents were measured with standard whole-cell, transferred for about 10 s to either a control saline solution con patch-clamp methods, as described elsewhere (Assad et al., 1989; taining 4 mM Ca:‘, or a saline solution buffered to 10 ’ M Ca-’ Assad and Corey, 1992). The internal solution contained Cs’ to with 5 mM BAPTA. The two samples were then transterred 10 block most of the current through Ca2+-activated K’ channels. normal saline solution containing 4 mM Ca” and mounted next Composition was 85 mM CsCI, 2 mM MgCI,, 10 mM EGTA, 2 mM to each other on a glass toverslip that had been coated threr Na2ATP, and 5 mM HEPES (pH 7.25). The external solution was a times with Cell-Tak (Collaborative Research. Inc.). Specimen5 normal frog saline, with the Caz+ concentration slightly elevated were then imrnedlately prepared for SEM. The sacculi were fixed and Cs’ added to block inwardly rectifying K’ channels. Com- for 60 min with 2% glutaraldehydr (Trd Pella, Inc.) in a buffer position was 120 mM NaCl, 2 mM KCI, 4 mM CaC12, 5 mM CsCI. containing 80 mM sodium cacodylate and 4 mM CaC&. All step\ 3 mM dextrose, and 5 mM HEPES (pH 7.25). The series resistance of the fixation were done on ice. In \omr case\, $acculi were of the patch pipette was routinely compensated by the patch transferred tor the final 20 min of the Inc-ubation Into a solution clamp (Yale Mk V with 1 Gn headstage), so that residual resis- containing the \ame concentration ot glutaraldehyde, plus 2’~~ tance was less than 4-6 MR. Stimuli were generated and rc- Triton X-100 detergent, to sotubitize the cell membranes. After sponses recorded with a PDP II/73 computer equipped with an several rinse5 with buffer, all tissues were postfixed for 30 mln INDEC interface. with 1% 050; tied Pelta, Inc.) In the same buffer. The saccull Mechanical Stimulation were again rinsed with buffer at room temperature and dehy Single hair bundles were moved along their morphologlcal axis drated in an ethanol series. The specimens were critical-point with a two-dimensional piezoelectric bimorph stimulator, of the dried from liquid CO> and sputter-c-oated with gold-palladium “pi” configuration (Corey and Hudspeth, 1980). Cells tended to Microscopy settle on the chamber bottom sideways, so that the kinocilium Specimens were observed on d field emlsblon scdnnlng rlectrc,n was at the far left or right edge of a bundle, as viewed from the microscope (Amray 1860FE, Hitachi S-4000, or JEOL. 6300F). Thr top; only these cells were studied. For most experiments, the sacculi were orlented so that the bevel of the hair bundles could glass stimulus probe was fabricated from a hollow pipette of be viewed en face. The macular epithetium was rapidly scanned about 0.7 pm tip diameter. Suction was applied to the back end at low power (about 3000x 1 in order to locate bundles that werr of the pipette through a fine polyethylene tube (Holton and undamaged and optimally oriented. tmages were photographed Hudspeth, 1986); under constant suction, the bulb of the kinoci- directly or rec.orded onto optical memory disk via the micrc lium adhered tightly to the tip of the stimulus probe. The rise scopes’ RS-170 video output\. time of the stimulator, measured with a photocell in an image Analysis of SFM Data plane of the microscope, was less than 1.5 ms (IO%-90%). Photographs or video prints ot the bundle\ wer$x coded as eitht,l Pressure Application of Low Cap BAPTA-treated or control, the code obscured, and the entire set Ca2+ was buffered to approximately 10~’ M with the Cal’ chela- of pictures numbered at random. Analysis was performed blind tor BAPTA (1,2-bis(o-aminophenoxy)ethane-N,N,N’,N’-tetraace- by four observers. Each observer was instructed to record the tic acid; Tsien, 1980). Except when noted, the buffer solution total number of tip links visible in each photograph, excludtng contained 5 mM BAPTA, 120 mM NaCI, 2 mM KCI, 0.1 mM CaCI,, broken or partially obscured tip links. The data were expressed 3 mM dextrose, and 5 mM HEPES. The solution was delivered by as a fraction of the total number of stereociliary tips for which solenoid-controlled pressure ejection from a pipette (ejection tip links could have been seen in each photograph and averaged pressure, *IO kPa). among the four observers. Because tips of adjacent stereocilia Hair Cell Transduction 991

were occasionally fused, this procedure underestimated the per- References centage of tip links present in each bundle. Assad, J. A., and Corey, D. P. (1992). An active motor mediates Transmission Electron Microscopy adaptation by vertebrate hair cells. J. Neurosci., in press. Preparation and Fixation Assad, J. A., Hacohen, N., and Corey, D. P. (1989). Voltage depen- Maculae were prepared and treated with Ca2+ or BAPTA as de- dence of adaptation and active bundle movement in bullfrog scribed for SEM. Five maculae were treated with 4 mM Ca2+ and saccular hair cells. Proc. Natl. Acad. Sci. USA 86, 2918-2922. 3 with 5 mM BAPTA. Samples were transferred to test and control Bagger-SjBbLk, D., and Werslll, J. (1973). The sensory hairs and solutions for 3-4 s and then returned to saline containing 4 mM of the basilar papilla in the lizard Calotes Ca*+ for 60 s. Maculae were fixed by immersion for 1 hr in 1% versicolor. J. Neurocytol. 2, 329-350. glutaraldehydefled Pella, Inc.) in saline buffered to pH 7.25with 20 mM HEPES. This and subsequent steps, except when noted, Corey, D. P., and Hudspeth, A. J. (1979a). Response latency of were performed at room temperature. Maculae were next vertebrate hair cells. Biophys. J. 26, 499-506. washed in 100 mM cacodylate buffer (pH 6.3). To enhance preser- Corey, D. P., and Hudspeth, A. J. (197913). Ionic basis of the recep vation and visualization of the actin cytoskeleton of stereocilia, tor potential in a vertebrate hair cell. Nature 287, 675-677. maculae were treated for 15 min with 5 PM phalloidin (Calbio- Corey, D. P., and Hudspeth, A. J. (1980). Mechanical stimulation them) in cacodylate buffer containing 0.1% Triton X-100. The and micromanipulation with piezoelectric bimorph elements. J. wash step was repeated, followed by postfixation in 1% OsOa Neurosci. Meth. 3, 183-202. (Ted Pella, Inc.) in 0.1 M phosphate buffer, for 30 min on ice. Corey, D. P., and Hudspeth, A. J. (1983). Kinetics of the receptor Maculae were washed thoroughly with distilled HZ0 and stained current in bullfrog saccular hair cells. J. Neurosci. 3, 962-976. en bloc with 0.5% uranyl acetate for 3 hr. After dehydration in a Crawford, A. C., Evans, M. C., and Fettiplace, R. (1989). Activation methanol series, samples were embedded in Spurr’s embedding and adaptation of transducer currents in turtle hair cells. J. Phys- plastic (Polysciences, Inc.), which was polymerized for 24 hr at iol. 479, 405-434. 65°C. Approximately 75 thin sections, cut parallel to the saccular nerve, were collected from each macula onto EM grids and Crawford, A. C., Evans, M. G., and Fettiplace, R. (1991). The ac- stained with saturated uranyl acetate and 0.2% lead citrate. tions of calcium on the mechano-electrical transducer current Microscopy of turtle hair cells. J. Physiol. 434, 369-398. Specimens were examined using a JEOL JEM IOOCX-II electron Eatock, R. A. Corey, D. P., and Hudspeth, A. J. (1987). Adaptation microscope operated at an accelerating voltage of 80 kV. Thin of mechanoelectrical transduction in hair cells of the bullfrog’s sections were systematically examined such that every seem- sacculus. J. Neurosci. 7, 2821-2836. ingly well-oriented bundle where a tip link might be seen was Forge, A., Davies, S., and Zajic, C. (1988). Characteristics of the photographed at low (6,700x or 10,000x) and higher (20,000x) membrane of the stereocilia and cell apex in cochlear hair cells. magnification. J. Neurocytol. 77, 325-334. Analysis of TFM Data Hacohen, N., Assad, J. A., Smith, W., and Corey, D. P. (1989). The criterion for evaluating the disposition of tip links was that Regulation of tension on hair-cell transduction channels: dis- onlythose pairs of stereocilia inwhich both sites oftip-link inser- placement and calcium dependence. J. Neurosci. I), 3988-3997. tion could be discerned were included in the analysis. These insertions are evident as pronounced osmophilic densities at the Hess, P., Lansman, J. B., andTsien, R. W. (1986). Calcium channel tip of the lower and side of the taller stereocilia. After these selectivity for divalent and monovalent cations. J. Gen. Physiol. optimally sectioned stereociliary pairs were selected, negatives 88, 293-319. and prints were coded as either BAPTA-treated or control, the Holton, T., and Hudspeth, A. J. (1986). The transduction channel code was obscured, and theentire set of pictures was numbered of hair cells from the bull-frog characterized by noise analysis. at random. Analysis was performed blind by four observers in- J. Physiol. 375, 195-227. structed to record the presence or absence of tip links in each Howard, J., and Hudspeth, A. J. (1987). Mechanical relaxation of selected stereociliary pair, using the negativeor positive images the hair bundle mediates adaptation in mechanoelectrical trans- as desired. In the same way, observers were instructed to score duction by the bullfrog’s saccular hair cell. Proc. Natl. Acad. Sci. the presence of tenting, as defined in Results. The number of USA 84, 3064-3068. tip links was expressed as a proportion of the total number of Howard, J., and Hudspeth, A. J. (1988). Compliance of the hair stereociliary pairs for which both densities were apparent and bundle associated with gating of the mechanoelrctrical trans- was averaged among the observers. duction channels in the bullfrog’s sacculal hair cell. Neuron 7, 189-199. Acknowledgments Howard, J., Roberts, W. M., and Hudspeth, A. J. (11988). Mecha- noelectrical transduction by hair cells. Annu. Rev. Biophys. Bio- We thank Drs. Bruce Bean, Steven Block, and Jonathon Howard phys. Chem. 77, 99-124. for comments on the manuscript, Robin Pinto and Karen Rock for electron microscopy, and Susan Cronin for laboratoryadmin- Hudspeth, A. J. (1982). Extracellular current flow and the site of istration. We are especially indebted to Dr. Bechara Kachar and transduction by vertebrate hair cells. J. Neurosci. 2, I-IO. Marianne Parakkal for assistance with transmission electron mi- Hudspeth, A. J. (1989). How the ’s works work. Nature 347, croscopy and use of their microscope facility at the National 397404. Institutes of Health, for the experiment in Figure 5. This work Jacobs, R. A., and Hudspeth, A. 1. (1990). Ultrastructural corre- was supported by grants from the NIH (DC-00304) and the Office lates of mechanoelectrical transduction in hair cells of the bull- of Naval Research (NOO14-91-J-1159), by the Howard Hughes frog’s internal ear. Cold Spring Harbor Symp. Quant. Biol. 55, Medical Institute, and by an NSF predoctoral fellowship to 547-561. J. A. A. D. P. C. is an Associate Investigator of the Howard Hughes Jaramillo, F., and Hudspeth, A. J. (1991). Localization of the hair Medical Institute, and J. A. A and G. M. G. S. are Ryan Fellows cell’s transduction channels at the hair bundle’s top1 by iontopho- at Harvard Medical School. retie application of a channel blocker. Neuron 7, 409-420. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be Jessell, T. M. (1988). Adhesion molecules and the’ hierarchy of hereby marked “advertisement in accordance with 18 USC Sec- neural development. Neuron 7, 3-13. tion 1734 solely to indicate this fact. Kimura, R. S. (1966). Hairs of the cochlear sensory cells and their attachment to the tectorial membrane. Acta Otolaryngol. 67, Received August 1, 1991; revised September 13, 1991. 55-72. NellrCOl 994

Ohmori, H. (1988). Mechanical stimulation and fura- fluores- cence in the hair bundle of dissociated hair cells of the chick. J. Physiol. 399, 115-137. Osborne, M. P., and Comis, S. D. (1990). Action of elastase, col- lagenase and other enzymes upon linkages between stereocilia in the guinea-pig . Acta Otolaryngol. 770, 37-45. Pickles, J. O., Comis, S. D., and Osborne, M. P. (1984). Cross-links between stereocilia in the guinea pig , and their possible relation to sensory transduction. Res. 75. 103- 112. Pickles,J. O., Osborne, M. P., and Comis, S. D. (1987). Vulnerabil- ity of tip links between stereocilia to acoustic trauma rn the guinea pig. Hearing Res. 25, 173-183. Pickles, J. O., Brix, J., Comis, 5. D., Cleich, O., Koppl, C., Manley, G. A., and Osborne, M. P. (1989). The organization of tip links and stereocilia on hair cells of bird and lizard basilar papillae. Hearing Res. 41, 31-42. Pickles, J. O., Brix, J., and Manley, C. A. (1990). Influence of colla- genase on tip links in hair cells of the chick basilar papilla. Hear- ing Res. 50, 139-143. Roberts, W. M., Jacobs, R. A., and Hudspeth, A. J. (1990). Colocali- zation of ion channels involved in frequency selectivity and syn- aptic transmission at presynaptic active zones of hair cells. J. Neurosci. 70, 3664-3684. Sand, 0. (1975). Effects of different ionic environments on the mechanosensitivity of lateral line organs in the mudpuppy. J. Comp. Physiol. 702, 27-42. Tsien, R. Y. (1980). New calcium indicators and buffers with high selectivity against magnesium and protons: design, synthesis, and properties of prototype structures. Biochemistry 79, 2396- 2404.